This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Update Win32 from version 0.54 to 0.56
[perl5.git] / pp_pack.c
index 17f7182..7a02b55 100644 (file)
--- a/pp_pack.c
+++ b/pp_pack.c
@@ -45,29 +45,29 @@ typedef struct tempsym {
   const char*    grpbeg;   /* 1st char of ()-group  */
   const char*    grpend;   /* end of ()-group       */
   I32      code;     /* template code (!<>)   */
-  I32      length;   /* length/repeat count   */
-  howlen_t howlen;   /* how length is given   */ 
-  int      level;    /* () nesting level      */
   U32      flags;    /* /=4, comma=2, pack=1  */
                      /*   and group modifiers */
+  SSize_t  length;   /* length/repeat count   */
+  howlen_t howlen;   /* how length is given   */ 
+  int      level;    /* () nesting level      */
   STRLEN   strbeg;   /* offset of group start */
   struct tempsym *previous; /* previous group */
 } tempsym_t;
 
 #define TEMPSYM_INIT(symptr, p, e, f) \
     STMT_START {       \
-       (symptr)->patptr   = (p);       \
-       (symptr)->patend   = (e);       \
-       (symptr)->grpbeg   = NULL;      \
-       (symptr)->grpend   = NULL;      \
-       (symptr)->grpend   = NULL;      \
-       (symptr)->code     = 0;         \
-       (symptr)->length   = 0;         \
-       (symptr)->howlen   = e_no_len;  \
-       (symptr)->level    = 0;         \
-       (symptr)->flags    = (f);       \
-       (symptr)->strbeg   = 0;         \
-       (symptr)->previous = NULL;      \
+        (symptr)->patptr   = (p);      \
+        (symptr)->patend   = (e);      \
+        (symptr)->grpbeg   = NULL;     \
+        (symptr)->grpend   = NULL;     \
+        (symptr)->grpend   = NULL;     \
+        (symptr)->code     = 0;                \
+        (symptr)->length   = 0;                \
+        (symptr)->howlen   = e_no_len; \
+        (symptr)->level    = 0;                \
+        (symptr)->flags    = (f);      \
+        (symptr)->strbeg   = 0;                \
+        (symptr)->previous = NULL;     \
    } STMT_END
 
 typedef union {
@@ -75,7 +75,7 @@ typedef union {
     U8 bytes[sizeof(NV)];
 } NV_bytes;
 
-#if defined(HAS_LONG_DOUBLE) && defined(USE_LONG_DOUBLE)
+#if defined(HAS_LONG_DOUBLE)
 typedef union {
     long double ld;
     U8 bytes[sizeof(long double)];
@@ -112,21 +112,17 @@ typedef union {
 /* CROSSCOMPILE and MULTIARCH are going to affect pp_pack() and pp_unpack().
    --jhi Feb 1999 */
 
-#if U16SIZE > SIZE16 || U32SIZE > SIZE32
-#  if BYTEORDER == 0x1234 || BYTEORDER == 0x12345678    /* little-endian */
-#    define OFF16(p)   ((char*)(p))
-#    define OFF32(p)   ((char*)(p))
-#  else
-#    if BYTEORDER == 0x4321 || BYTEORDER == 0x87654321  /* big-endian */
-#      define OFF16(p) ((char*)(p) + (sizeof(U16) - SIZE16))
-#      define OFF32(p) ((char*)(p) + (sizeof(U32) - SIZE32))
-#    else
-       ++++ bad cray byte order
-#    endif
-#  endif
-#else
+#if U16SIZE <= SIZE16 && U32SIZE <= SIZE32
 #  define OFF16(p)     ((char *) (p))
 #  define OFF32(p)     ((char *) (p))
+#elif BYTEORDER == 0x1234 || BYTEORDER == 0x12345678    /* little-endian */
+#  define OFF16(p)     ((char*)(p))
+#  define OFF32(p)     ((char*)(p))
+#elif BYTEORDER == 0x4321 || BYTEORDER == 0x87654321  /* big-endian */
+#  define OFF16(p)     ((char*)(p) + (sizeof(U16) - SIZE16))
+#  define OFF32(p)     ((char*)(p) + (sizeof(U32) - SIZE32))
+#else
+#  error "bad cray byte order"
 #endif
 
 #define PUSH16(utf8, cur, p, needs_swap)                        \
@@ -142,7 +138,7 @@ typedef union {
 #  error "Unsupported byteorder"
         /* Need to add code here to re-instate mixed endian support.
            NEEDS_SWAP would need to hold a flag indicating which action to
-           take, and S_reverse_copy and the code in uni_to_bytes would need
+           take, and S_reverse_copy and the code in S_utf8_to_bytes would need
            logic adding to deal with any mixed-endian transformations needed.
         */
 #endif
@@ -151,8 +147,8 @@ typedef union {
 #define SHIFT_BYTES(utf8, s, strend, buf, len, datumtype, needs_swap)  \
 STMT_START {                                           \
     if (UNLIKELY(utf8)) {                               \
-        if (!uni_to_bytes(aTHX_ &s, strend,            \
-         (char *) (buf), len, datumtype)) break;       \
+        if (!S_utf8_to_bytes(aTHX_ &s, strend,         \
+          (char *) (buf), len, datumtype)) break;      \
     } else {                                           \
         if (UNLIKELY(needs_swap))                       \
             S_reverse_copy(s, (char *) (buf), len);     \
@@ -195,7 +191,7 @@ S_mul128(pTHX_ SV *sv, U8 m)
 
   PERL_ARGS_ASSERT_MUL128;
 
-  if (!strnEQ(s, "0000", 4)) {  /* need to grow sv */
+  if (! memBEGINs(s, len, "0000")) {  /* need to grow sv */
     SV * const tmpNew = newSVpvs("0000000000");
 
     sv_catsv(tmpNew, sv);
@@ -216,16 +212,9 @@ S_mul128(pTHX_ SV *sv, U8 m)
 
 /* Explosives and implosives. */
 
-#if 'I' == 73 && 'J' == 74
-/* On an ASCII/ISO kind of system */
-#define ISUUCHAR(ch)    ((ch) >= ' ' && (ch) < 'a')
-#else
-/*
-  Some other sort of character set - use memchr() so we don't match
-  the null byte.
- */
-#define ISUUCHAR(ch)    (memchr(PL_uuemap, (ch), sizeof(PL_uuemap)-1) || (ch) == ' ')
-#endif
+#define ISUUCHAR(ch)    inRANGE(NATIVE_TO_LATIN1(ch),               \
+                                NATIVE_TO_LATIN1(' '),              \
+                                NATIVE_TO_LATIN1('a') - 1)
 
 /* type modifiers */
 #define TYPE_IS_SHRIEKING      0x100
@@ -245,7 +234,7 @@ S_mul128(pTHX_ SV *sv, U8 m)
 #define PACK_SIZE_UNPREDICTABLE                0x40    /* Not a fixed size element */
 #define PACK_SIZE_MASK                 0x3F
 
-#include "packsizetables.c"
+#include "packsizetables.inc"
 
 static void
 S_reverse_copy(const char *src, char *dest, STRLEN len)
@@ -256,56 +245,59 @@ S_reverse_copy(const char *src, char *dest, STRLEN len)
 }
 
 STATIC U8
-uni_to_byte(pTHX_ const char **s, const char *end, I32 datumtype)
+utf8_to_byte(pTHX_ const char **s, const char *end, I32 datumtype)
 {
     STRLEN retlen;
-    UV val = utf8n_to_uvchr((U8 *) *s, end-*s, &retlen,
-                        ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
-    /* We try to process malformed UTF-8 as much as possible (preferably with
-       warnings), but these two mean we make no progress in the string and
-       might enter an infinite loop */
-    if (retlen == (STRLEN) -1 || retlen == 0)
-       Perl_croak(aTHX_ "Malformed UTF-8 string in '%c' format in unpack",
-                  (int) TYPE_NO_MODIFIERS(datumtype));
+    UV val;
+
+    if (*s >= end) {
+        goto croak;
+    }
+    val = utf8n_to_uvchr((U8 *) *s, end-*s, &retlen,
+                         ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
+    if (retlen == (STRLEN) -1)
+      croak:
+        Perl_croak(aTHX_ "Malformed UTF-8 string in '%c' format in unpack",
+                   (int) TYPE_NO_MODIFIERS(datumtype));
     if (val >= 0x100) {
-       Perl_ck_warner(aTHX_ packWARN(WARN_UNPACK),
-                      "Character in '%c' format wrapped in unpack",
-                      (int) TYPE_NO_MODIFIERS(datumtype));
-       val &= 0xff;
+        Perl_ck_warner(aTHX_ packWARN(WARN_UNPACK),
+                       "Character in '%c' format wrapped in unpack",
+                       (int) TYPE_NO_MODIFIERS(datumtype));
+        val &= 0xff;
     }
     *s += retlen;
     return (U8)val;
 }
 
 #define SHIFT_BYTE(utf8, s, strend, datumtype) ((utf8) ? \
-       uni_to_byte(aTHX_ &(s), (strend), (datumtype)) : \
-       *(U8 *)(s)++)
+        utf8_to_byte(aTHX_ &(s), (strend), (datumtype)) : \
+        *(U8 *)(s)++)
 
 STATIC bool
-uni_to_bytes(pTHX_ const char **s, const char *end, const char *buf, int buf_len, I32 datumtype)
+S_utf8_to_bytes(pTHX_ const char **s, const char *end, const char *buf, SSize_t buf_len, I32 datumtype)
 {
     UV val;
     STRLEN retlen;
     const char *from = *s;
     int bad = 0;
     const U32 flags = ckWARN(WARN_UTF8) ?
-       UTF8_CHECK_ONLY : (UTF8_CHECK_ONLY | UTF8_ALLOW_ANY);
+        UTF8_CHECK_ONLY : (UTF8_CHECK_ONLY | UTF8_ALLOW_ANY);
     const bool needs_swap = NEEDS_SWAP(datumtype);
 
     if (UNLIKELY(needs_swap))
         buf += buf_len;
 
     for (;buf_len > 0; buf_len--) {
-       if (from >= end) return FALSE;
-       val = utf8n_to_uvchr((U8 *) from, end-from, &retlen, flags);
-       if (retlen == (STRLEN) -1 || retlen == 0) {
-           from += UTF8SKIP(from);
-           bad |= 1;
-       } else from += retlen;
-       if (val >= 0x100) {
-           bad |= 2;
-           val &= 0xff;
-       }
+        if (from >= end) return FALSE;
+        val = utf8n_to_uvchr((U8 *) from, end-from, &retlen, flags);
+        if (retlen == (STRLEN) -1) {
+            from += UTF8_SAFE_SKIP(from, end);
+            bad |= 1;
+        } else from += retlen;
+        if (val >= 0x100) {
+            bad |= 2;
+            val &= 0xff;
+        }
         if (UNLIKELY(needs_swap))
             *(U8 *)--buf = (U8)val;
         else
@@ -313,45 +305,30 @@ uni_to_bytes(pTHX_ const char **s, const char *end, const char *buf, int buf_len
     }
     /* We have enough characters for the buffer. Did we have problems ? */
     if (bad) {
-       if (bad & 1) {
-           /* Rewalk the string fragment while warning */
-           const char *ptr;
-           const int flags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
-           for (ptr = *s; ptr < from; ptr += UTF8SKIP(ptr)) {
-               if (ptr >= end) break;
-               utf8n_to_uvchr((U8 *) ptr, end-ptr, &retlen, flags);
-           }
-           if (from > end) from = end;
-       }
-       if ((bad & 2))
-           Perl_ck_warner(aTHX_ packWARN(datumtype & TYPE_IS_PACK ?
-                                      WARN_PACK : WARN_UNPACK),
-                          "Character(s) in '%c' format wrapped in %s",
-                          (int) TYPE_NO_MODIFIERS(datumtype),
-                          datumtype & TYPE_IS_PACK ? "pack" : "unpack");
+        if (bad & 1) {
+            /* Rewalk the string fragment while warning */
+            const char *ptr;
+            const U32 flags = ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY;
+            for (ptr = *s; ptr < from; ptr += UTF8SKIP(ptr)) {
+                if (ptr >= end) break;
+                utf8n_to_uvchr((U8 *) ptr, end-ptr, &retlen, flags);
+            }
+            if (from > end) from = end;
+        }
+        if ((bad & 2))
+            Perl_ck_warner(aTHX_ packWARN(datumtype & TYPE_IS_PACK ?
+                                       WARN_PACK : WARN_UNPACK),
+                           "Character(s) in '%c' format wrapped in %s",
+                           (int) TYPE_NO_MODIFIERS(datumtype),
+                           datumtype & TYPE_IS_PACK ? "pack" : "unpack");
     }
     *s = from;
     return TRUE;
 }
 
-STATIC bool
-next_uni_uu(pTHX_ const char **s, const char *end, I32 *out)
-{
-    STRLEN retlen;
-    const UV val = utf8n_to_uvchr((U8 *) *s, end-*s, &retlen, UTF8_CHECK_ONLY);
-    if (val >= 0x100 || !ISUUCHAR(val) ||
-       retlen == (STRLEN) -1 || retlen == 0) {
-       *out = 0;
-       return FALSE;
-    }
-    *out = PL_uudmap[val] & 077;
-    *s += retlen;
-    return TRUE;
-}
-
 STATIC char *
-S_bytes_to_uni(const U8 *start, STRLEN len, char *dest, const bool needs_swap) {
-    PERL_ARGS_ASSERT_BYTES_TO_UNI;
+S_my_bytes_to_utf8(const U8 *start, STRLEN len, char *dest, const bool needs_swap) {
+    PERL_ARGS_ASSERT_MY_BYTES_TO_UTF8;
 
     if (UNLIKELY(needs_swap)) {
         const U8 *p = start + len;
@@ -371,23 +348,40 @@ S_bytes_to_uni(const U8 *start, STRLEN len, char *dest, const bool needs_swap) {
 #define PUSH_BYTES(utf8, cur, buf, len, needs_swap)             \
 STMT_START {                                                   \
     if (UNLIKELY(utf8))                                                \
-       (cur) = S_bytes_to_uni((U8 *) buf, len, (cur), needs_swap);       \
+        (cur) = my_bytes_to_utf8((U8 *) buf, len, (cur), needs_swap);       \
     else {                                                     \
         if (UNLIKELY(needs_swap))                               \
             S_reverse_copy((char *)(buf), cur, len);            \
         else                                                    \
             Copy(buf, cur, len, char);                         \
-       (cur) += (len);                                         \
+        (cur) += (len);                                                \
     }                                                          \
 } STMT_END
 
+#define SAFE_UTF8_EXPAND(var)  \
+STMT_START {                           \
+    if ((var) > SSize_t_MAX / UTF8_EXPAND) \
+        Perl_croak(aTHX_ "%s", "Out of memory during pack()"); \
+    (var) = (var) * UTF8_EXPAND; \
+} STMT_END
+
+#define GROWING2(utf8, cat, start, cur, item_size, item_count) \
+STMT_START {                                                   \
+    if (SSize_t_MAX / (item_size) < (item_count))              \
+        Perl_croak(aTHX_ "%s", "Out of memory during pack()"); \
+    GROWING((utf8), (cat), (start), (cur), (item_size) * (item_count)); \
+} STMT_END
+
 #define GROWING(utf8, cat, start, cur, in_len) \
 STMT_START {                                   \
     STRLEN glen = (in_len);                    \
-    if (utf8) glen *= UTF8_EXPAND;             \
-    if ((cur) + glen >= (start) + SvLEN(cat)) {        \
-       (start) = sv_exp_grow(cat, glen);       \
-       (cur) = (start) + SvCUR(cat);           \
+    STRLEN catcur = (STRLEN)((cur) - (start)); \
+    if (utf8) SAFE_UTF8_EXPAND(glen);          \
+    if (SSize_t_MAX - glen < catcur)           \
+        Perl_croak(aTHX_ "%s", "Out of memory during pack()"); \
+    if (catcur + glen >= SvLEN(cat)) { \
+        (start) = sv_exp_grow(cat, glen);      \
+        (cur) = (start) + SvCUR(cat);          \
     }                                          \
 } STMT_END
 
@@ -395,12 +389,12 @@ STMT_START {                                      \
 STMT_START {                                   \
     const STRLEN glen = (in_len);              \
     STRLEN gl = glen;                          \
-    if (utf8) gl *= UTF8_EXPAND;               \
+    if (utf8) SAFE_UTF8_EXPAND(gl);            \
     if ((cur) + gl >= (start) + SvLEN(cat)) {  \
         *cur = '\0';                           \
         SvCUR_set((cat), (cur) - (start));     \
-       (start) = sv_exp_grow(cat, gl);         \
-       (cur) = (start) + SvCUR(cat);           \
+        (start) = sv_exp_grow(cat, gl);                \
+        (cur) = (start) + SvCUR(cat);          \
     }                                          \
     PUSH_BYTES(utf8, cur, buf, glen, 0);        \
 } STMT_END
@@ -408,8 +402,8 @@ STMT_START {                                        \
 #define PUSH_BYTE(utf8, s, byte)               \
 STMT_START {                                   \
     if (utf8) {                                        \
-       const U8 au8 = (byte);                  \
-       (s) = S_bytes_to_uni(&au8, 1, (s), 0);  \
+        const U8 au8 = (byte);                 \
+        (s) = my_bytes_to_utf8(&au8, 1, (s), 0);\
     } else *(U8 *)(s)++ = (byte);              \
 } STMT_END
 
@@ -419,9 +413,9 @@ STMT_START {                                                        \
     STRLEN retlen;                                             \
     if (str >= end) break;                                     \
     val = utf8n_to_uvchr((U8 *) str, end-str, &retlen, utf8_flags);    \
-    if (retlen == (STRLEN) -1 || retlen == 0) {                        \
-       *cur = '\0';                                            \
-       Perl_croak(aTHX_ "Malformed UTF-8 string in pack");     \
+    if (retlen == (STRLEN) -1) {                               \
+        *cur = '\0';                                           \
+        Perl_croak(aTHX_ "Malformed UTF-8 string in pack");    \
     }                                                          \
     str += retlen;                                             \
 } STMT_END
@@ -432,109 +426,108 @@ static const char *_action( const tempsym_t* symptr )
 }
 
 /* Returns the sizeof() struct described by pat */
-STATIC I32
+STATIC SSize_t
 S_measure_struct(pTHX_ tempsym_t* symptr)
 {
-    I32 total = 0;
+    SSize_t total = 0;
 
     PERL_ARGS_ASSERT_MEASURE_STRUCT;
 
     while (next_symbol(symptr)) {
-       I32 len;
-       int size;
+        SSize_t len, size;
 
         switch (symptr->howlen) {
-         case e_star:
-           Perl_croak(aTHX_ "Within []-length '*' not allowed in %s",
+          case e_star:
+            Perl_croak(aTHX_ "Within []-length '*' not allowed in %s",
                         _action( symptr ) );
 
-         default:
-           /* e_no_len and e_number */
-           len = symptr->length;
-           break;
+          default:
+            /* e_no_len and e_number */
+            len = symptr->length;
+            break;
         }
 
-       size = packprops[TYPE_NO_ENDIANNESS(symptr->code)] & PACK_SIZE_MASK;
-       if (!size) {
-            int star;
-           /* endianness doesn't influence the size of a type */
-           switch(TYPE_NO_ENDIANNESS(symptr->code)) {
-           default:
-               Perl_croak(aTHX_ "Invalid type '%c' in %s",
-                          (int)TYPE_NO_MODIFIERS(symptr->code),
+        size = packprops[TYPE_NO_ENDIANNESS(symptr->code)] & PACK_SIZE_MASK;
+        if (!size) {
+            SSize_t star;
+            /* endianness doesn't influence the size of a type */
+            switch(TYPE_NO_ENDIANNESS(symptr->code)) {
+            default:
+                Perl_croak(aTHX_ "Invalid type '%c' in %s",
+                           (int)TYPE_NO_MODIFIERS(symptr->code),
                            _action( symptr ) );
-           case '.' | TYPE_IS_SHRIEKING:
-           case '@' | TYPE_IS_SHRIEKING:
-           case '@':
-           case '.':
-           case '/':
-           case 'U':                   /* XXXX Is it correct? */
-           case 'w':
-           case 'u':
-               Perl_croak(aTHX_ "Within []-length '%c' not allowed in %s",
-                          (int) TYPE_NO_MODIFIERS(symptr->code),
+            case '.' | TYPE_IS_SHRIEKING:
+            case '@' | TYPE_IS_SHRIEKING:
+            case '@':
+            case '.':
+            case '/':
+            case 'U':                  /* XXXX Is it correct? */
+            case 'w':
+            case 'u':
+                Perl_croak(aTHX_ "Within []-length '%c' not allowed in %s",
+                           (int) TYPE_NO_MODIFIERS(symptr->code),
                            _action( symptr ) );
-           case '%':
-               size = 0;
-               break;
-           case '(':
-           {
-               tempsym_t savsym = *symptr;
-               symptr->patptr = savsym.grpbeg;
-               symptr->patend = savsym.grpend;
-               /* XXXX Theoretically, we need to measure many times at
-                  different positions, since the subexpression may contain
-                  alignment commands, but be not of aligned length.
-                  Need to detect this and croak().  */
-               size = measure_struct(symptr);
-               *symptr = savsym;
-               break;
-           }
-           case 'X' | TYPE_IS_SHRIEKING:
-               /* XXXX Is this useful?  Then need to treat MEASURE_BACKWARDS.
-                */
-               if (!len)               /* Avoid division by 0 */
-                   len = 1;
-               len = total % len;      /* Assumed: the start is aligned. */
-               /* FALLTHROUGH */
-           case 'X':
-               size = -1;
-               if (total < len)
+            case '%':
+                size = 0;
+                break;
+            case '(':
+            {
+                tempsym_t savsym = *symptr;
+                symptr->patptr = savsym.grpbeg;
+                symptr->patend = savsym.grpend;
+                /* XXXX Theoretically, we need to measure many times at
+                   different positions, since the subexpression may contain
+                   alignment commands, but be not of aligned length.
+                   Need to detect this and croak().  */
+                size = measure_struct(symptr);
+                *symptr = savsym;
+                break;
+            }
+            case 'X' | TYPE_IS_SHRIEKING:
+                /* XXXX Is this useful?  Then need to treat MEASURE_BACKWARDS.
+                 */
+                if (!len)              /* Avoid division by 0 */
+                    len = 1;
+                len = total % len;     /* Assumed: the start is aligned. */
+                /* FALLTHROUGH */
+            case 'X':
+                size = -1;
+                if (total < len)
                     Perl_croak(aTHX_ "'X' outside of string in %s", _action( symptr ) );
-               break;
-           case 'x' | TYPE_IS_SHRIEKING:
-               if (!len)               /* Avoid division by 0 */
-                   len = 1;
-               star = total % len;     /* Assumed: the start is aligned. */
-               if (star)               /* Other portable ways? */
-                   len = len - star;
-               else
-                   len = 0;
-               /* FALLTHROUGH */
-           case 'x':
-           case 'A':
-           case 'Z':
-           case 'a':
-               size = 1;
-               break;
-           case 'B':
-           case 'b':
-               len = (len + 7)/8;
-               size = 1;
-               break;
-           case 'H':
-           case 'h':
-               len = (len + 1)/2;
-               size = 1;
-               break;
-
-           case 'P':
-               len = 1;
-               size = sizeof(char*);
-               break;
-           }
-       }
-       total += len * size;
+                break;
+            case 'x' | TYPE_IS_SHRIEKING:
+                if (!len)              /* Avoid division by 0 */
+                    len = 1;
+                star = total % len;    /* Assumed: the start is aligned. */
+                if (star)              /* Other portable ways? */
+                    len = len - star;
+                else
+                    len = 0;
+                /* FALLTHROUGH */
+            case 'x':
+            case 'A':
+            case 'Z':
+            case 'a':
+                size = 1;
+                break;
+            case 'B':
+            case 'b':
+                len = (len + 7)/8;
+                size = 1;
+                break;
+            case 'H':
+            case 'h':
+                len = (len + 1)/2;
+                size = 1;
+                break;
+
+            case 'P':
+                len = 1;
+                size = sizeof(char*);
+                break;
+            }
+        }
+        total += len * size;
     }
     return total;
 }
@@ -549,20 +542,20 @@ S_group_end(pTHX_ const char *patptr, const char *patend, char ender)
     PERL_ARGS_ASSERT_GROUP_END;
 
     while (patptr < patend) {
-       const char c = *patptr++;
-
-       if (isSPACE(c))
-           continue;
-       else if (c == ender)
-           return patptr-1;
-       else if (c == '#') {
-           while (patptr < patend && *patptr != '\n')
-               patptr++;
-           continue;
-       } else if (c == '(')
-           patptr = group_end(patptr, patend, ')') + 1;
-       else if (c == '[')
-           patptr = group_end(patptr, patend, ']') + 1;
+        const char c = *patptr++;
+
+        if (isSPACE(c))
+            continue;
+        else if (c == ender)
+            return patptr-1;
+        else if (c == '#') {
+            while (patptr < patend && *patptr != '\n')
+                patptr++;
+            continue;
+        } else if (c == '(')
+            patptr = group_end(patptr, patend, ')') + 1;
+        else if (c == '[')
+            patptr = group_end(patptr, patend, ']') + 1;
     }
     Perl_croak(aTHX_ "No group ending character '%c' found in template",
                ender);
@@ -575,16 +568,17 @@ S_group_end(pTHX_ const char *patptr, const char *patend, char ender)
  * Advances char pointer to 1st non-digit char and returns number
  */
 STATIC const char *
-S_get_num(pTHX_ const char *patptr, I32 *lenptr )
+S_get_num(pTHX_ const char *patptr, SSize_t *lenptr )
 {
-  I32 len = *patptr++ - '0';
+  SSize_t len = *patptr++ - '0';
 
   PERL_ARGS_ASSERT_GET_NUM;
 
   while (isDIGIT(*patptr)) {
-    if (len >= 0x7FFFFFFF/10)
+    SSize_t nlen = (len * 10) + (*patptr++ - '0');
+    if (nlen < 0 || nlen/10 != len)
       Perl_croak(aTHX_ "pack/unpack repeat count overflow");
-    len = (len * 10) + (*patptr++ - '0');
+    len = nlen;
   }
   *lenptr = len;
   return patptr;
@@ -609,21 +603,21 @@ S_next_symbol(pTHX_ tempsym_t* symptr )
     else if (*patptr == '#') {
       patptr++;
       while (patptr < patend && *patptr != '\n')
-       patptr++;
+        patptr++;
       if (patptr < patend)
-       patptr++;
+        patptr++;
     } else {
       /* We should have found a template code */
       I32 code = *patptr++ & 0xFF;
       U32 inherited_modifiers = 0;
 
       if (code == ','){ /* grandfather in commas but with a warning */
-       if (((symptr->flags & FLAG_COMMA) == 0) && ckWARN(WARN_UNPACK)){
+        if (((symptr->flags & FLAG_COMMA) == 0) && ckWARN(WARN_UNPACK)){
           symptr->flags |= FLAG_COMMA;
-         Perl_warner(aTHX_ packWARN(WARN_UNPACK),
-                     "Invalid type ',' in %s", _action( symptr ) );
+          Perl_warner(aTHX_ packWARN(WARN_UNPACK),
+                      "Invalid type ',' in %s", _action( symptr ) );
         }
-       continue;
+        continue;
       }
 
       /* for '(', skip to ')' */
@@ -634,7 +628,7 @@ S_next_symbol(pTHX_ tempsym_t* symptr )
         symptr->grpbeg = patptr;
         patptr = 1 + ( symptr->grpend = group_end(patptr, patend, ')') );
         if( symptr->level >= MAX_SUB_TEMPLATE_LEVEL )
-         Perl_croak(aTHX_ "Too deeply nested ()-groups in %s",
+          Perl_croak(aTHX_ "Too deeply nested ()-groups in %s",
                         _action( symptr ) );
       }
 
@@ -683,10 +677,10 @@ S_next_symbol(pTHX_ tempsym_t* symptr )
                      *patptr, _action( symptr ) );
 
         if ((code & modifier)) {
-           Perl_ck_warner(aTHX_ packWARN(WARN_UNPACK),
-                          "Duplicate modifier '%c' after '%c' in %s",
-                          *patptr, (int) TYPE_NO_MODIFIERS(code),
-                          _action( symptr ) );
+            Perl_ck_warner(aTHX_ packWARN(WARN_UNPACK),
+                           "Duplicate modifier '%c' after '%c' in %s",
+                           *patptr, (int) TYPE_NO_MODIFIERS(code),
+                           _action( symptr ) );
         }
 
         code |= modifier;
@@ -698,8 +692,8 @@ S_next_symbol(pTHX_ tempsym_t* symptr )
 
       /* look for count and/or / */
       if (patptr < patend) {
-       if (isDIGIT(*patptr)) {
-         patptr = get_num( patptr, &symptr->length );
+        if (isDIGIT(*patptr)) {
+          patptr = get_num( patptr, &symptr->length );
           symptr->howlen = e_number;
 
         } else if (*patptr == '*') {
@@ -735,9 +729,9 @@ S_next_symbol(pTHX_ tempsym_t* symptr )
           else if (*patptr == '#') {
             patptr++;
             while (patptr < patend && *patptr != '\n')
-             patptr++;
+              patptr++;
             if (patptr < patend)
-             patptr++;
+              patptr++;
           } else {
             if (*patptr == '/') {
               symptr->flags |= FLAG_SLASH;
@@ -748,8 +742,8 @@ S_next_symbol(pTHX_ tempsym_t* symptr )
                             _action( symptr ) );
             }
             break;
-         }
-       }
+          }
+        }
       } else {
         /* at end - no count, no / */
         symptr->howlen = e_no_len;
@@ -782,14 +776,14 @@ need_utf8(const char *pat, const char *patend)
     PERL_ARGS_ASSERT_NEED_UTF8;
 
     while (pat < patend) {
-       if (pat[0] == '#') {
-           pat++;
-           pat = (const char *) memchr(pat, '\n', patend-pat);
-           if (!pat) return FALSE;
-       } else if (pat[0] == 'U') {
-           if (first || pat[1] == '0') return TRUE;
-       } else first = FALSE;
-       pat++;
+        if (pat[0] == '#') {
+            pat++;
+            pat = (const char *) memchr(pat, '\n', patend-pat);
+            if (!pat) return FALSE;
+        } else if (pat[0] == 'U') {
+            if (first || pat[1] == '0') return TRUE;
+        } else first = FALSE;
+        pat++;
     }
     return FALSE;
 }
@@ -799,40 +793,38 @@ first_symbol(const char *pat, const char *patend) {
     PERL_ARGS_ASSERT_FIRST_SYMBOL;
 
     while (pat < patend) {
-       if (pat[0] != '#') return pat[0];
-       pat++;
-       pat = (const char *) memchr(pat, '\n', patend-pat);
-       if (!pat) return 0;
-       pat++;
+        if (pat[0] != '#') return pat[0];
+        pat++;
+        pat = (const char *) memchr(pat, '\n', patend-pat);
+        if (!pat) return 0;
+        pat++;
     }
     return 0;
 }
 
 /*
 
-=head1 Pack and Unpack
-
 =for apidoc unpackstring
 
-The engine implementing the unpack() Perl function.
+The engine implementing the C<unpack()> Perl function.
 
-Using the template pat..patend, this function unpacks the string
-s..strend into a number of mortal SVs, which it pushes onto the perl
-argument (@_) stack (so you will need to issue a C<PUTBACK> before and
+Using the template C<pat..patend>, this function unpacks the string
+C<s..strend> into a number of mortal SVs, which it pushes onto the perl
+argument (C<@_>) stack (so you will need to issue a C<PUTBACK> before and
 C<SPAGAIN> after the call to this function).  It returns the number of
 pushed elements.
 
-The strend and patend pointers should point to the byte following the last
-character of each string.
+The C<strend> and C<patend> pointers should point to the byte following the
+last character of each string.
 
 Although this function returns its values on the perl argument stack, it
 doesn't take any parameters from that stack (and thus in particular
-there's no need to do a PUSHMARK before calling it, unlike L</call_pv> for
+there's no need to do a C<PUSHMARK> before calling it, unlike L</call_pv> for
 example).
 
 =cut */
 
-I32
+SSize_t
 Perl_unpackstring(pTHX_ const char *pat, const char *patend, const char *s, const char *strend, U32 flags)
 {
     tempsym_t sym;
@@ -841,34 +833,34 @@ Perl_unpackstring(pTHX_ const char *pat, const char *patend, const char *s, cons
 
     if (flags & FLAG_DO_UTF8) flags |= FLAG_WAS_UTF8;
     else if (need_utf8(pat, patend)) {
-       /* We probably should try to avoid this in case a scalar context call
-          wouldn't get to the "U0" */
-       STRLEN len = strend - s;
-       s = (char *) bytes_to_utf8((U8 *) s, &len);
-       SAVEFREEPV(s);
-       strend = s + len;
-       flags |= FLAG_DO_UTF8;
+        /* We probably should try to avoid this in case a scalar context call
+           wouldn't get to the "U0" */
+        STRLEN len = strend - s;
+        s = (char *) bytes_to_utf8((U8 *) s, &len);
+        SAVEFREEPV(s);
+        strend = s + len;
+        flags |= FLAG_DO_UTF8;
     }
 
     if (first_symbol(pat, patend) != 'U' && (flags & FLAG_DO_UTF8))
-       flags |= FLAG_PARSE_UTF8;
+        flags |= FLAG_PARSE_UTF8;
 
     TEMPSYM_INIT(&sym, pat, patend, flags);
 
     return unpack_rec(&sym, s, s, strend, NULL );
 }
 
-STATIC I32
+STATIC SSize_t
 S_unpack_rec(pTHX_ tempsym_t* symptr, const char *s, const char *strbeg, const char *strend, const char **new_s )
 {
     dSP;
     SV *sv = NULL;
-    const I32 start_sp_offset = SP - PL_stack_base;
+    const SSize_t start_sp_offset = SP - PL_stack_base;
     howlen_t howlen;
-    I32 checksum = 0;
+    SSize_t checksum = 0;
     UV cuv = 0;
     NV cdouble = 0.0;
-    const int bits_in_uv = CHAR_BIT * sizeof(cuv);
+    const SSize_t bits_in_uv = CHAR_BIT * sizeof(cuv);
     bool beyond = FALSE;
     bool explicit_length;
     const bool unpack_only_one = (symptr->flags & FLAG_UNPACK_ONLY_ONE) != 0;
@@ -879,823 +871,835 @@ S_unpack_rec(pTHX_ tempsym_t* symptr, const char *s, const char *strbeg, const c
     symptr->strbeg = s - strbeg;
 
     while (next_symbol(symptr)) {
-       packprops_t props;
-       I32 len;
+        packprops_t props;
+        SSize_t len;
         I32 datumtype = symptr->code;
         bool needs_swap;
-       /* do first one only unless in list context
-          / is implemented by unpacking the count, then popping it from the
-          stack, so must check that we're not in the middle of a /  */
+        /* do first one only unless in list context
+           / is implemented by unpacking the count, then popping it from the
+           stack, so must check that we're not in the middle of a /  */
         if ( unpack_only_one
-            && (SP - PL_stack_base == start_sp_offset + 1)
-            && (datumtype != '/') )   /* XXX can this be omitted */
+             && (SP - PL_stack_base == start_sp_offset + 1)
+             && (datumtype != '/') )   /* XXX can this be omitted */
             break;
 
         switch (howlen = symptr->howlen) {
-         case e_star:
-           len = strend - strbeg;      /* long enough */
-           break;
-         default:
-           /* e_no_len and e_number */
-           len = symptr->length;
-           break;
+          case e_star:
+            len = strend - strbeg;     /* long enough */
+            break;
+          default:
+            /* e_no_len and e_number */
+            len = symptr->length;
+            break;
         }
 
         explicit_length = TRUE;
       redo_switch:
         beyond = s >= strend;
 
-       props = packprops[TYPE_NO_ENDIANNESS(datumtype)];
-       if (props) {
-           /* props nonzero means we can process this letter. */
-            const long size = props & PACK_SIZE_MASK;
-            const long howmany = (strend - s) / size;
-           if (len > howmany)
-               len = howmany;
-
-           if (!checksum || (props & PACK_SIZE_CANNOT_CSUM)) {
-               if (len && unpack_only_one) len = 1;
-               EXTEND(SP, len);
-               EXTEND_MORTAL(len);
-           }
-       }
+        props = packprops[TYPE_NO_ENDIANNESS(datumtype)];
+        if (props) {
+            /* props nonzero means we can process this letter. */
+            const SSize_t size = props & PACK_SIZE_MASK;
+            const SSize_t howmany = (strend - s) / size;
+            if (len > howmany)
+                len = howmany;
+
+            if (!checksum || (props & PACK_SIZE_CANNOT_CSUM)) {
+                if (len && unpack_only_one) len = 1;
+                EXTEND(SP, len);
+                EXTEND_MORTAL(len);
+            }
+        }
 
         needs_swap = NEEDS_SWAP(datumtype);
 
-       switch(TYPE_NO_ENDIANNESS(datumtype)) {
-       default:
-           Perl_croak(aTHX_ "Invalid type '%c' in unpack", (int)TYPE_NO_MODIFIERS(datumtype) );
+        switch(TYPE_NO_ENDIANNESS(datumtype)) {
+        default:
+            Perl_croak(aTHX_ "Invalid type '%c' in unpack", (int)TYPE_NO_MODIFIERS(datumtype) );
 
-       case '%':
-           if (howlen == e_no_len)
-               len = 16;               /* len is not specified */
-           checksum = len;
-           cuv = 0;
-           cdouble = 0;
-           continue;
+        case '%':
+            if (howlen == e_no_len)
+                len = 16;              /* len is not specified */
+            checksum = len;
+            cuv = 0;
+            cdouble = 0;
+            continue;
 
-       case '(':
-       {
+        case '(':
+        {
             tempsym_t savsym = *symptr;
             const U32 group_modifiers = TYPE_MODIFIERS(datumtype & ~symptr->flags);
-           symptr->flags |= group_modifiers;
+            symptr->flags |= group_modifiers;
             symptr->patend = savsym.grpend;
-           symptr->previous = &savsym;
+            symptr->previous = &savsym;
             symptr->level++;
-           PUTBACK;
-           if (len && unpack_only_one) len = 1;
-           while (len--) {
-               symptr->patptr = savsym.grpbeg;
-               if (utf8) symptr->flags |=  FLAG_PARSE_UTF8;
-               else      symptr->flags &= ~FLAG_PARSE_UTF8;
-               unpack_rec(symptr, s, strbeg, strend, &s);
+            PUTBACK;
+            if (len && unpack_only_one) len = 1;
+            while (len--) {
+                symptr->patptr = savsym.grpbeg;
+                if (utf8) symptr->flags |=  FLAG_PARSE_UTF8;
+                else      symptr->flags &= ~FLAG_PARSE_UTF8;
+                unpack_rec(symptr, s, strbeg, strend, &s);
                 if (s == strend && savsym.howlen == e_star)
-                   break; /* No way to continue */
-           }
-           SPAGAIN;
+                    break; /* No way to continue */
+            }
+            SPAGAIN;
             savsym.flags = symptr->flags & ~group_modifiers;
             *symptr = savsym;
-           break;
-       }
-       case '.' | TYPE_IS_SHRIEKING:
-       case '.': {
-           const char *from;
-           SV *sv;
-           const bool u8 = utf8 && !(datumtype & TYPE_IS_SHRIEKING);
-           if (howlen == e_star) from = strbeg;
-           else if (len <= 0) from = s;
-           else {
-               tempsym_t *group = symptr;
-
-               while (--len && group) group = group->previous;
-               from = group ? strbeg + group->strbeg : strbeg;
-           }
-           sv = from <= s ?
-               newSVuv(  u8 ? (UV) utf8_length((const U8*)from, (const U8*)s) : (UV) (s-from)) :
-               newSViv(-(u8 ? (IV) utf8_length((const U8*)s, (const U8*)from) : (IV) (from-s)));
-           mXPUSHs(sv);
-           break;
-       }
-       case '@' | TYPE_IS_SHRIEKING:
-       case '@':
-           s = strbeg + symptr->strbeg;
-           if (utf8  && !(datumtype & TYPE_IS_SHRIEKING))
-           {
-               while (len > 0) {
-                   if (s >= strend)
-                       Perl_croak(aTHX_ "'@' outside of string in unpack");
-                   s += UTF8SKIP(s);
-                   len--;
-               }
-               if (s > strend)
-                   Perl_croak(aTHX_ "'@' outside of string with malformed UTF-8 in unpack");
-           } else {
-               if (strend-s < len)
-                   Perl_croak(aTHX_ "'@' outside of string in unpack");
-               s += len;
-           }
-           break;
-       case 'X' | TYPE_IS_SHRIEKING:
-           if (!len)                   /* Avoid division by 0 */
-               len = 1;
-           if (utf8) {
-               const char *hop, *last;
-               I32 l = len;
-               hop = last = strbeg;
-               while (hop < s) {
-                   hop += UTF8SKIP(hop);
-                   if (--l == 0) {
-                       last = hop;
-                       l = len;
-                   }
-               }
-               if (last > s)
-                   Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
-               s = last;
-               break;
-           }
-           len = (s - strbeg) % len;
-           /* FALLTHROUGH */
-       case 'X':
-           if (utf8) {
-               while (len > 0) {
-                   if (s <= strbeg)
-                       Perl_croak(aTHX_ "'X' outside of string in unpack");
-                   while (--s, UTF8_IS_CONTINUATION(*s)) {
-                       if (s <= strbeg)
-                           Perl_croak(aTHX_ "'X' outside of string in unpack");
-                   }
-                   len--;
-               }
-           } else {
-               if (len > s - strbeg)
-                   Perl_croak(aTHX_ "'X' outside of string in unpack" );
-               s -= len;
-           }
-           break;
-       case 'x' | TYPE_IS_SHRIEKING: {
-            I32 ai32;
-           if (!len)                   /* Avoid division by 0 */
-               len = 1;
-           if (utf8) ai32 = utf8_length((U8 *) strbeg, (U8 *) s) % len;
-           else      ai32 = (s - strbeg)                         % len;
-           if (ai32 == 0) break;
-           len -= ai32;
-            }
-           /* FALLTHROUGH */
-       case 'x':
-           if (utf8) {
-               while (len>0) {
-                   if (s >= strend)
-                       Perl_croak(aTHX_ "'x' outside of string in unpack");
-                   s += UTF8SKIP(s);
-                   len--;
-               }
-           } else {
-               if (len > strend - s)
-                   Perl_croak(aTHX_ "'x' outside of string in unpack");
-               s += len;
-           }
-           break;
-       case '/':
-           Perl_croak(aTHX_ "'/' must follow a numeric type in unpack");
-
-       case 'A':
-       case 'Z':
-       case 'a':
-           if (checksum) {
-               /* Preliminary length estimate is assumed done in 'W' */
-               if (len > strend - s) len = strend - s;
-               goto W_checksum;
-           }
-           if (utf8) {
-               I32 l;
-               const char *hop;
-               for (l=len, hop=s; l>0; l--, hop += UTF8SKIP(hop)) {
-                   if (hop >= strend) {
-                       if (hop > strend)
-                           Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
-                       break;
-                   }
-               }
-               if (hop > strend)
-                   Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
-               len = hop - s;
-           } else if (len > strend - s)
-               len = strend - s;
-
-           if (datumtype == 'Z') {
-               /* 'Z' strips stuff after first null */
-               const char *ptr, *end;
-               end = s + len;
-               for (ptr = s; ptr < end; ptr++) if (*ptr == 0) break;
-               sv = newSVpvn(s, ptr-s);
-               if (howlen == e_star) /* exact for 'Z*' */
-                   len = ptr-s + (ptr != strend ? 1 : 0);
-           } else if (datumtype == 'A') {
-               /* 'A' strips both nulls and spaces */
-               const char *ptr;
-               if (utf8 && (symptr->flags & FLAG_WAS_UTF8)) {
-                   for (ptr = s+len-1; ptr >= s; ptr--)
-                       if (*ptr != 0 && !UTF8_IS_CONTINUATION(*ptr) &&
-                           !isSPACE_utf8(ptr)) break;
-                   if (ptr >= s) ptr += UTF8SKIP(ptr);
-                   else ptr++;
-                   if (ptr > s+len)
-                       Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
-               } else {
-                   for (ptr = s+len-1; ptr >= s; ptr--)
-                       if (*ptr != 0 && !isSPACE(*ptr)) break;
-                   ptr++;
-               }
-               sv = newSVpvn(s, ptr-s);
-           } else sv = newSVpvn(s, len);
-
-           if (utf8) {
-               SvUTF8_on(sv);
-               /* Undo any upgrade done due to need_utf8() */
-               if (!(symptr->flags & FLAG_WAS_UTF8))
-                   sv_utf8_downgrade(sv, 0);
-           }
-           mXPUSHs(sv);
-           s += len;
-           break;
-       case 'B':
-       case 'b': {
-           char *str;
-           if (howlen == e_star || len > (strend - s) * 8)
-               len = (strend - s) * 8;
-           if (checksum) {
-               if (utf8)
-                   while (len >= 8 && s < strend) {
-                       cuv += PL_bitcount[uni_to_byte(aTHX_ &s, strend, datumtype)];
-                       len -= 8;
-                   }
-               else
-                   while (len >= 8) {
-                       cuv += PL_bitcount[*(U8 *)s++];
-                       len -= 8;
-                   }
-               if (len && s < strend) {
-                   U8 bits;
-                   bits = SHIFT_BYTE(utf8, s, strend, datumtype);
-                   if (datumtype == 'b')
-                       while (len-- > 0) {
-                           if (bits & 1) cuv++;
-                           bits >>= 1;
-                       }
-                   else
-                       while (len-- > 0) {
-                           if (bits & 0x80) cuv++;
-                           bits <<= 1;
-                       }
-               }
-               break;
-           }
-
-           sv = sv_2mortal(newSV(len ? len : 1));
-           SvPOK_on(sv);
-           str = SvPVX(sv);
-           if (datumtype == 'b') {
-               U8 bits = 0;
-               const I32 ai32 = len;
-               for (len = 0; len < ai32; len++) {
-                   if (len & 7) bits >>= 1;
-                   else if (utf8) {
-                       if (s >= strend) break;
-                       bits = uni_to_byte(aTHX_ &s, strend, datumtype);
-                   } else bits = *(U8 *) s++;
-                   *str++ = bits & 1 ? '1' : '0';
-               }
-           } else {
-               U8 bits = 0;
-               const I32 ai32 = len;
-               for (len = 0; len < ai32; len++) {
-                   if (len & 7) bits <<= 1;
-                   else if (utf8) {
-                       if (s >= strend) break;
-                       bits = uni_to_byte(aTHX_ &s, strend, datumtype);
-                   } else bits = *(U8 *) s++;
-                   *str++ = bits & 0x80 ? '1' : '0';
-               }
-           }
-           *str = '\0';
-           SvCUR_set(sv, str - SvPVX_const(sv));
-           XPUSHs(sv);
-           break;
-       }
-       case 'H':
-       case 'h': {
-           char *str = NULL;
-           /* Preliminary length estimate, acceptable for utf8 too */
-           if (howlen == e_star || len > (strend - s) * 2)
-               len = (strend - s) * 2;
-           if (!checksum) {
-               sv = sv_2mortal(newSV(len ? len : 1));
-               SvPOK_on(sv);
-               str = SvPVX(sv);
-           }
-           if (datumtype == 'h') {
-               U8 bits = 0;
-               I32 ai32 = len;
-               for (len = 0; len < ai32; len++) {
-                   if (len & 1) bits >>= 4;
-                   else if (utf8) {
-                       if (s >= strend) break;
-                       bits = uni_to_byte(aTHX_ &s, strend, datumtype);
-                   } else bits = * (U8 *) s++;
-                   if (!checksum)
-                       *str++ = PL_hexdigit[bits & 15];
-               }
-           } else {
-               U8 bits = 0;
-               const I32 ai32 = len;
-               for (len = 0; len < ai32; len++) {
-                   if (len & 1) bits <<= 4;
-                   else if (utf8) {
-                       if (s >= strend) break;
-                       bits = uni_to_byte(aTHX_ &s, strend, datumtype);
-                   } else bits = *(U8 *) s++;
-                   if (!checksum)
-                       *str++ = PL_hexdigit[(bits >> 4) & 15];
-               }
-           }
-           if (!checksum) {
-               *str = '\0';
-               SvCUR_set(sv, str - SvPVX_const(sv));
-               XPUSHs(sv);
-           }
-           break;
-       }
-       case 'C':
+            break;
+        }
+        case '.' | TYPE_IS_SHRIEKING:
+        case '.': {
+            const char *from;
+            SV *sv;
+            const bool u8 = utf8 && !(datumtype & TYPE_IS_SHRIEKING);
+            if (howlen == e_star) from = strbeg;
+            else if (len <= 0) from = s;
+            else {
+                tempsym_t *group = symptr;
+
+                while (--len && group) group = group->previous;
+                from = group ? strbeg + group->strbeg : strbeg;
+            }
+            sv = from <= s ?
+                newSVuv(  u8 ? (UV) utf8_length((const U8*)from, (const U8*)s) : (UV) (s-from)) :
+                newSViv(-(u8 ? (IV) utf8_length((const U8*)s, (const U8*)from) : (IV) (from-s)));
+            mXPUSHs(sv);
+            break;
+        }
+        case '@' | TYPE_IS_SHRIEKING:
+        case '@':
+            s = strbeg + symptr->strbeg;
+            if (utf8  && !(datumtype & TYPE_IS_SHRIEKING))
+            {
+                while (len > 0) {
+                    if (s >= strend)
+                        Perl_croak(aTHX_ "'@' outside of string in unpack");
+                    s += UTF8SKIP(s);
+                    len--;
+                }
+                if (s > strend)
+                    Perl_croak(aTHX_ "'@' outside of string with malformed UTF-8 in unpack");
+            } else {
+                if (strend-s < len)
+                    Perl_croak(aTHX_ "'@' outside of string in unpack");
+                s += len;
+            }
+            break;
+        case 'X' | TYPE_IS_SHRIEKING:
+            if (!len)                  /* Avoid division by 0 */
+                len = 1;
+            if (utf8) {
+                const char *hop, *last;
+                SSize_t l = len;
+                hop = last = strbeg;
+                while (hop < s) {
+                    hop += UTF8SKIP(hop);
+                    if (--l == 0) {
+                        last = hop;
+                        l = len;
+                    }
+                }
+                if (last > s)
+                    Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
+                s = last;
+                break;
+            }
+            len = (s - strbeg) % len;
+            /* FALLTHROUGH */
+        case 'X':
+            if (utf8) {
+                while (len > 0) {
+                    if (s <= strbeg)
+                        Perl_croak(aTHX_ "'X' outside of string in unpack");
+                    while (--s, UTF8_IS_CONTINUATION(*s)) {
+                        if (s <= strbeg)
+                            Perl_croak(aTHX_ "'X' outside of string in unpack");
+                    }
+                    len--;
+                }
+            } else {
+                if (len > s - strbeg)
+                    Perl_croak(aTHX_ "'X' outside of string in unpack" );
+                s -= len;
+            }
+            break;
+        case 'x' | TYPE_IS_SHRIEKING: {
+            SSize_t ai32;
+            if (!len)                  /* Avoid division by 0 */
+                len = 1;
+            if (utf8) ai32 = utf8_length((U8 *) strbeg, (U8 *) s) % len;
+            else      ai32 = (s - strbeg)                         % len;
+            if (ai32 == 0) break;
+            len -= ai32;
+            }
+            /* FALLTHROUGH */
+        case 'x':
+            if (utf8) {
+                while (len>0) {
+                    if (s >= strend)
+                        Perl_croak(aTHX_ "'x' outside of string in unpack");
+                    s += UTF8SKIP(s);
+                    len--;
+                }
+            } else {
+                if (len > strend - s)
+                    Perl_croak(aTHX_ "'x' outside of string in unpack");
+                s += len;
+            }
+            break;
+        case '/':
+            Perl_croak(aTHX_ "'/' must follow a numeric type in unpack");
+
+        case 'A':
+        case 'Z':
+        case 'a':
+            if (checksum) {
+                /* Preliminary length estimate is assumed done in 'W' */
+                if (len > strend - s) len = strend - s;
+                goto W_checksum;
+            }
+            if (utf8) {
+                SSize_t l;
+                const char *hop;
+                for (l=len, hop=s; l>0; l--, hop += UTF8SKIP(hop)) {
+                    if (hop >= strend) {
+                        if (hop > strend)
+                            Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
+                        break;
+                    }
+                }
+                if (hop > strend)
+                    Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
+                len = hop - s;
+            } else if (len > strend - s)
+                len = strend - s;
+
+            if (datumtype == 'Z') {
+                /* 'Z' strips stuff after first null */
+                const char *ptr, *end;
+                end = s + len;
+                for (ptr = s; ptr < end; ptr++) if (*ptr == 0) break;
+                sv = newSVpvn(s, ptr-s);
+                if (howlen == e_star) /* exact for 'Z*' */
+                    len = ptr-s + (ptr != strend ? 1 : 0);
+            } else if (datumtype == 'A') {
+                /* 'A' strips both nulls and spaces */
+                const char *ptr;
+                if (utf8 && (symptr->flags & FLAG_WAS_UTF8)) {
+                    for (ptr = s+len-1; ptr >= s; ptr--) {
+                        if (   *ptr != 0
+                            && !UTF8_IS_CONTINUATION(*ptr)
+                            && !isSPACE_utf8_safe(ptr, strend))
+                        {
+                            break;
+                        }
+                    }
+                    if (ptr >= s) ptr += UTF8SKIP(ptr);
+                    else ptr++;
+                    if (ptr > s+len)
+                        Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
+                } else {
+                    for (ptr = s+len-1; ptr >= s; ptr--)
+                        if (*ptr != 0 && !isSPACE(*ptr)) break;
+                    ptr++;
+                }
+                sv = newSVpvn(s, ptr-s);
+            } else sv = newSVpvn(s, len);
+
+            if (utf8) {
+                SvUTF8_on(sv);
+                /* Undo any upgrade done due to need_utf8() */
+                if (!(symptr->flags & FLAG_WAS_UTF8))
+                    sv_utf8_downgrade(sv, 0);
+            }
+            mXPUSHs(sv);
+            s += len;
+            break;
+        case 'B':
+        case 'b': {
+            char *str;
+            if (howlen == e_star || len > (strend - s) * 8)
+                len = (strend - s) * 8;
+            if (checksum) {
+                if (utf8)
+                    while (len >= 8 && s < strend) {
+                        cuv += PL_bitcount[utf8_to_byte(aTHX_ &s, strend, datumtype)];
+                        len -= 8;
+                    }
+                else
+                    while (len >= 8) {
+                        cuv += PL_bitcount[*(U8 *)s++];
+                        len -= 8;
+                    }
+                if (len && s < strend) {
+                    U8 bits;
+                    bits = SHIFT_BYTE(utf8, s, strend, datumtype);
+                    if (datumtype == 'b')
+                        while (len-- > 0) {
+                            if (bits & 1) cuv++;
+                            bits >>= 1;
+                        }
+                    else
+                        while (len-- > 0) {
+                            if (bits & 0x80) cuv++;
+                            bits <<= 1;
+                        }
+                }
+                break;
+            }
+
+            sv = sv_2mortal(newSV(len ? len : 1));
+            SvPOK_on(sv);
+            str = SvPVX(sv);
+            if (datumtype == 'b') {
+                U8 bits = 0;
+                const SSize_t ai32 = len;
+                for (len = 0; len < ai32; len++) {
+                    if (len & 7) bits >>= 1;
+                    else if (utf8) {
+                        if (s >= strend) break;
+                        bits = utf8_to_byte(aTHX_ &s, strend, datumtype);
+                    } else bits = *(U8 *) s++;
+                    *str++ = bits & 1 ? '1' : '0';
+                }
+            } else {
+                U8 bits = 0;
+                const SSize_t ai32 = len;
+                for (len = 0; len < ai32; len++) {
+                    if (len & 7) bits <<= 1;
+                    else if (utf8) {
+                        if (s >= strend) break;
+                        bits = utf8_to_byte(aTHX_ &s, strend, datumtype);
+                    } else bits = *(U8 *) s++;
+                    *str++ = bits & 0x80 ? '1' : '0';
+                }
+            }
+            *str = '\0';
+            SvCUR_set(sv, str - SvPVX_const(sv));
+            XPUSHs(sv);
+            break;
+        }
+        case 'H':
+        case 'h': {
+            char *str = NULL;
+            /* Preliminary length estimate, acceptable for utf8 too */
+            if (howlen == e_star || len > (strend - s) * 2)
+                len = (strend - s) * 2;
+            if (!checksum) {
+                sv = sv_2mortal(newSV(len ? len : 1));
+                SvPOK_on(sv);
+                str = SvPVX(sv);
+            }
+            if (datumtype == 'h') {
+                U8 bits = 0;
+                SSize_t ai32 = len;
+                for (len = 0; len < ai32; len++) {
+                    if (len & 1) bits >>= 4;
+                    else if (utf8) {
+                        if (s >= strend) break;
+                        bits = utf8_to_byte(aTHX_ &s, strend, datumtype);
+                    } else bits = * (U8 *) s++;
+                    if (!checksum)
+                        *str++ = PL_hexdigit[bits & 15];
+                }
+            } else {
+                U8 bits = 0;
+                const SSize_t ai32 = len;
+                for (len = 0; len < ai32; len++) {
+                    if (len & 1) bits <<= 4;
+                    else if (utf8) {
+                        if (s >= strend) break;
+                        bits = utf8_to_byte(aTHX_ &s, strend, datumtype);
+                    } else bits = *(U8 *) s++;
+                    if (!checksum)
+                        *str++ = PL_hexdigit[(bits >> 4) & 15];
+                }
+            }
+            if (!checksum) {
+                *str = '\0';
+                SvCUR_set(sv, str - SvPVX_const(sv));
+                XPUSHs(sv);
+            }
+            break;
+        }
+        case 'C':
             if (len == 0) {
                 if (explicit_length)
-                   /* Switch to "character" mode */
-                   utf8 = (symptr->flags & FLAG_DO_UTF8) ? 1 : 0;
-               break;
-           }
-           /* FALLTHROUGH */
-       case 'c':
-           while (len-- > 0 && s < strend) {
-               int aint;
-               if (utf8)
-                 {
-                   STRLEN retlen;
-                   aint = utf8n_to_uvchr((U8 *) s, strend-s, &retlen,
-                                ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
-                   if (retlen == (STRLEN) -1 || retlen == 0)
-                       Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
-                   s += retlen;
-                 }
-               else
-                 aint = *(U8 *)(s)++;
-               if (aint >= 128 && datumtype != 'C')    /* fake up signed chars */
-                   aint -= 256;
-               if (!checksum)
-                   mPUSHi(aint);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)aint;
-               else
-                   cuv += aint;
-           }
-           break;
-       case 'W':
-         W_checksum:
-           if (utf8) {
-               while (len-- > 0 && s < strend) {
-                   STRLEN retlen;
-                   const UV val = utf8n_to_uvchr((U8 *) s, strend-s, &retlen,
-                                        ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
-                   if (retlen == (STRLEN) -1 || retlen == 0)
-                       Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
-                   s += retlen;
-                   if (!checksum)
-                       mPUSHu(val);
-                   else if (checksum > bits_in_uv)
-                       cdouble += (NV) val;
-                   else
-                       cuv += val;
-               }
-           } else if (!checksum)
-               while (len-- > 0) {
-                   const U8 ch = *(U8 *) s++;
-                   mPUSHu(ch);
-           }
-           else if (checksum > bits_in_uv)
-               while (len-- > 0) cdouble += (NV) *(U8 *) s++;
-           else
-               while (len-- > 0) cuv += *(U8 *) s++;
-           break;
-       case 'U':
-           if (len == 0) {
+                    /* Switch to "character" mode */
+                    utf8 = (symptr->flags & FLAG_DO_UTF8) ? 1 : 0;
+                break;
+            }
+            /* FALLTHROUGH */
+        case 'c':
+            while (len-- > 0 && s < strend) {
+                int aint;
+                if (utf8)
+                  {
+                    STRLEN retlen;
+                    aint = utf8n_to_uvchr((U8 *) s, strend-s, &retlen,
+                                 ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
+                    if (retlen == (STRLEN) -1)
+                        Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
+                    s += retlen;
+                  }
+                else
+                  aint = *(U8 *)(s)++;
+                if (aint >= 128 && datumtype != 'C')   /* fake up signed chars */
+                    aint -= 256;
+                if (!checksum)
+                    mPUSHi(aint);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)aint;
+                else
+                    cuv += aint;
+            }
+            break;
+        case 'W':
+          W_checksum:
+            if (utf8) {
+                while (len-- > 0 && s < strend) {
+                    STRLEN retlen;
+                    const UV val = utf8n_to_uvchr((U8 *) s, strend-s, &retlen,
+                                         ckWARN(WARN_UTF8) ? 0 : UTF8_ALLOW_ANY);
+                    if (retlen == (STRLEN) -1)
+                        Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
+                    s += retlen;
+                    if (!checksum)
+                        mPUSHu(val);
+                    else if (checksum > bits_in_uv)
+                        cdouble += (NV) val;
+                    else
+                        cuv += val;
+                }
+            } else if (!checksum)
+                while (len-- > 0) {
+                    const U8 ch = *(U8 *) s++;
+                    mPUSHu(ch);
+            }
+            else if (checksum > bits_in_uv)
+                while (len-- > 0) cdouble += (NV) *(U8 *) s++;
+            else
+                while (len-- > 0) cuv += *(U8 *) s++;
+            break;
+        case 'U':
+            if (len == 0) {
                 if (explicit_length && howlen != e_star) {
-                   /* Switch to "bytes in UTF-8" mode */
-                   if (symptr->flags & FLAG_DO_UTF8) utf8 = 0;
-                   else
-                       /* Should be impossible due to the need_utf8() test */
-                       Perl_croak(aTHX_ "U0 mode on a byte string");
-               }
-               break;
-           }
-           if (len > strend - s) len = strend - s;
-           if (!checksum) {
-               if (len && unpack_only_one) len = 1;
-               EXTEND(SP, len);
-               EXTEND_MORTAL(len);
-           }
-           while (len-- > 0 && s < strend) {
-               STRLEN retlen;
-               UV auv;
-               if (utf8) {
-                   U8 result[UTF8_MAXLEN];
-                   const char *ptr = s;
-                   STRLEN len;
-                   /* Bug: warns about bad utf8 even if we are short on bytes
-                      and will break out of the loop */
-                   if (!uni_to_bytes(aTHX_ &ptr, strend, (char *) result, 1,
-                                     'U'))
-                       break;
-                   len = UTF8SKIP(result);
-                   if (!uni_to_bytes(aTHX_ &ptr, strend,
-                                     (char *) &result[1], len-1, 'U')) break;
-                   auv = utf8n_to_uvchr(result, len, &retlen, UTF8_ALLOW_DEFAULT);
-                   s = ptr;
-               } else {
-                   auv = utf8n_to_uvchr((U8*)s, strend - s, &retlen, UTF8_ALLOW_DEFAULT);
-                   if (retlen == (STRLEN) -1 || retlen == 0)
-                       Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
-                   s += retlen;
-               }
-               if (!checksum)
-                   mPUSHu(auv);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV) auv;
-               else
-                   cuv += auv;
-           }
-           break;
-       case 's' | TYPE_IS_SHRIEKING:
+                    /* Switch to "bytes in UTF-8" mode */
+                    if (symptr->flags & FLAG_DO_UTF8) utf8 = 0;
+                    else
+                        /* Should be impossible due to the need_utf8() test */
+                        Perl_croak(aTHX_ "U0 mode on a byte string");
+                }
+                break;
+            }
+            if (len > strend - s) len = strend - s;
+            if (!checksum) {
+                if (len && unpack_only_one) len = 1;
+                EXTEND(SP, len);
+                EXTEND_MORTAL(len);
+            }
+            while (len-- > 0 && s < strend) {
+                STRLEN retlen;
+                UV auv;
+                if (utf8) {
+                    U8 result[UTF8_MAXLEN+1];
+                    const char *ptr = s;
+                    STRLEN len;
+                    /* Bug: warns about bad utf8 even if we are short on bytes
+                       and will break out of the loop */
+                    if (!S_utf8_to_bytes(aTHX_ &ptr, strend, (char *) result, 1,
+                                      'U'))
+                        break;
+                    len = UTF8SKIP(result);
+                    if (!S_utf8_to_bytes(aTHX_ &ptr, strend,
+                                      (char *) &result[1], len-1, 'U')) break;
+                    auv = NATIVE_TO_UNI(utf8n_to_uvchr(result,
+                                                       len,
+                                                       &retlen,
+                                                       UTF8_ALLOW_DEFAULT));
+                    s = ptr;
+                } else {
+                    auv = NATIVE_TO_UNI(utf8n_to_uvchr((U8*)s,
+                                                       strend - s,
+                                                       &retlen,
+                                                       UTF8_ALLOW_DEFAULT));
+                    if (retlen == (STRLEN) -1)
+                        Perl_croak(aTHX_ "Malformed UTF-8 string in unpack");
+                    s += retlen;
+                }
+                if (!checksum)
+                    mPUSHu(auv);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV) auv;
+                else
+                    cuv += auv;
+            }
+            break;
+        case 's' | TYPE_IS_SHRIEKING:
 #if SHORTSIZE != SIZE16
-           while (len-- > 0) {
-               short ashort;
+            while (len-- > 0) {
+                short ashort;
                 SHIFT_VAR(utf8, s, strend, ashort, datumtype, needs_swap);
-               if (!checksum)
-                   mPUSHi(ashort);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)ashort;
-               else
-                   cuv += ashort;
-           }
-           break;
+                if (!checksum)
+                    mPUSHi(ashort);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)ashort;
+                else
+                    cuv += ashort;
+            }
+            break;
 #else
-           /* FALLTHROUGH */
+            /* FALLTHROUGH */
 #endif
-       case 's':
-           while (len-- > 0) {
-               I16 ai16;
+        case 's':
+            while (len-- > 0) {
+                I16 ai16;
 
 #if U16SIZE > SIZE16
-               ai16 = 0;
+                ai16 = 0;
 #endif
                 SHIFT16(utf8, s, strend, &ai16, datumtype, needs_swap);
 #if U16SIZE > SIZE16
-               if (ai16 > 32767)
-                   ai16 -= 65536;
+                if (ai16 > 32767)
+                    ai16 -= 65536;
 #endif
-               if (!checksum)
-                   mPUSHi(ai16);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)ai16;
-               else
-                   cuv += ai16;
-           }
-           break;
-       case 'S' | TYPE_IS_SHRIEKING:
+                if (!checksum)
+                    mPUSHi(ai16);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)ai16;
+                else
+                    cuv += ai16;
+            }
+            break;
+        case 'S' | TYPE_IS_SHRIEKING:
 #if SHORTSIZE != SIZE16
-           while (len-- > 0) {
-               unsigned short aushort;
-                SHIFT_VAR(utf8, s, strend, aushort, datumtype, needs_swap,
-                          needs_swap);
-               if (!checksum)
-                   mPUSHu(aushort);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)aushort;
-               else
-                   cuv += aushort;
-           }
-           break;
+            while (len-- > 0) {
+                unsigned short aushort;
+                SHIFT_VAR(utf8, s, strend, aushort, datumtype, needs_swap);
+                if (!checksum)
+                    mPUSHu(aushort);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)aushort;
+                else
+                    cuv += aushort;
+            }
+            break;
 #else
             /* FALLTHROUGH */
 #endif
-       case 'v':
-       case 'n':
-       case 'S':
-           while (len-- > 0) {
-               U16 au16;
+        case 'v':
+        case 'n':
+        case 'S':
+            while (len-- > 0) {
+                U16 au16;
 #if U16SIZE > SIZE16
-               au16 = 0;
+                au16 = 0;
 #endif
                 SHIFT16(utf8, s, strend, &au16, datumtype, needs_swap);
-               if (datumtype == 'n')
-                   au16 = PerlSock_ntohs(au16);
-               if (datumtype == 'v')
-                   au16 = vtohs(au16);
-               if (!checksum)
-                   mPUSHu(au16);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV) au16;
-               else
-                   cuv += au16;
-           }
-           break;
-       case 'v' | TYPE_IS_SHRIEKING:
-       case 'n' | TYPE_IS_SHRIEKING:
-           while (len-- > 0) {
-               I16 ai16;
+                if (datumtype == 'n')
+                    au16 = PerlSock_ntohs(au16);
+                if (datumtype == 'v')
+                    au16 = vtohs(au16);
+                if (!checksum)
+                    mPUSHu(au16);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV) au16;
+                else
+                    cuv += au16;
+            }
+            break;
+        case 'v' | TYPE_IS_SHRIEKING:
+        case 'n' | TYPE_IS_SHRIEKING:
+            while (len-- > 0) {
+                I16 ai16;
 # if U16SIZE > SIZE16
-               ai16 = 0;
+                ai16 = 0;
 # endif
                 SHIFT16(utf8, s, strend, &ai16, datumtype, needs_swap);
                 /* There should never be any byte-swapping here.  */
                 assert(!TYPE_ENDIANNESS(datumtype));
-               if (datumtype == ('n' | TYPE_IS_SHRIEKING))
-                   ai16 = (I16) PerlSock_ntohs((U16) ai16);
-               if (datumtype == ('v' | TYPE_IS_SHRIEKING))
-                   ai16 = (I16) vtohs((U16) ai16);
-               if (!checksum)
-                   mPUSHi(ai16);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV) ai16;
-               else
-                   cuv += ai16;
-           }
-           break;
-       case 'i':
-       case 'i' | TYPE_IS_SHRIEKING:
-           while (len-- > 0) {
-               int aint;
+                if (datumtype == ('n' | TYPE_IS_SHRIEKING))
+                    ai16 = (I16) PerlSock_ntohs((U16) ai16);
+                if (datumtype == ('v' | TYPE_IS_SHRIEKING))
+                    ai16 = (I16) vtohs((U16) ai16);
+                if (!checksum)
+                    mPUSHi(ai16);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV) ai16;
+                else
+                    cuv += ai16;
+            }
+            break;
+        case 'i':
+        case 'i' | TYPE_IS_SHRIEKING:
+            while (len-- > 0) {
+                int aint;
                 SHIFT_VAR(utf8, s, strend, aint, datumtype, needs_swap);
-               if (!checksum)
-                   mPUSHi(aint);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)aint;
-               else
-                   cuv += aint;
-           }
-           break;
-       case 'I':
-       case 'I' | TYPE_IS_SHRIEKING:
-           while (len-- > 0) {
-               unsigned int auint;
+                if (!checksum)
+                    mPUSHi(aint);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)aint;
+                else
+                    cuv += aint;
+            }
+            break;
+        case 'I':
+        case 'I' | TYPE_IS_SHRIEKING:
+            while (len-- > 0) {
+                unsigned int auint;
                 SHIFT_VAR(utf8, s, strend, auint, datumtype, needs_swap);
-               if (!checksum)
-                   mPUSHu(auint);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)auint;
-               else
-                   cuv += auint;
-           }
-           break;
-       case 'j':
-           while (len-- > 0) {
-               IV aiv;
+                if (!checksum)
+                    mPUSHu(auint);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)auint;
+                else
+                    cuv += auint;
+            }
+            break;
+        case 'j':
+            while (len-- > 0) {
+                IV aiv;
                 SHIFT_VAR(utf8, s, strend, aiv, datumtype, needs_swap);
-               if (!checksum)
-                   mPUSHi(aiv);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)aiv;
-               else
-                   cuv += aiv;
-           }
-           break;
-       case 'J':
-           while (len-- > 0) {
-               UV auv;
+                if (!checksum)
+                    mPUSHi(aiv);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)aiv;
+                else
+                    cuv += aiv;
+            }
+            break;
+        case 'J':
+            while (len-- > 0) {
+                UV auv;
                 SHIFT_VAR(utf8, s, strend, auv, datumtype, needs_swap);
-               if (!checksum)
-                   mPUSHu(auv);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)auv;
-               else
-                   cuv += auv;
-           }
-           break;
-       case 'l' | TYPE_IS_SHRIEKING:
+                if (!checksum)
+                    mPUSHu(auv);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)auv;
+                else
+                    cuv += auv;
+            }
+            break;
+        case 'l' | TYPE_IS_SHRIEKING:
 #if LONGSIZE != SIZE32
-           while (len-- > 0) {
-               long along;
+            while (len-- > 0) {
+                long along;
                 SHIFT_VAR(utf8, s, strend, along, datumtype, needs_swap);
-               if (!checksum)
-                   mPUSHi(along);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)along;
-               else
-                   cuv += along;
-           }
-           break;
+                if (!checksum)
+                    mPUSHi(along);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)along;
+                else
+                    cuv += along;
+            }
+            break;
 #else
-           /* FALLTHROUGH */
+            /* FALLTHROUGH */
 #endif
-       case 'l':
-           while (len-- > 0) {
-               I32 ai32;
+        case 'l':
+            while (len-- > 0) {
+                I32 ai32;
 #if U32SIZE > SIZE32
-               ai32 = 0;
+                ai32 = 0;
 #endif
                 SHIFT32(utf8, s, strend, &ai32, datumtype, needs_swap);
 #if U32SIZE > SIZE32
-               if (ai32 > 2147483647) ai32 -= 4294967296;
+                if (ai32 > 2147483647) ai32 -= 4294967296;
 #endif
-               if (!checksum)
-                   mPUSHi(ai32);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)ai32;
-               else
-                   cuv += ai32;
-           }
-           break;
-       case 'L' | TYPE_IS_SHRIEKING:
+                if (!checksum)
+                    mPUSHi(ai32);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)ai32;
+                else
+                    cuv += ai32;
+            }
+            break;
+        case 'L' | TYPE_IS_SHRIEKING:
 #if LONGSIZE != SIZE32
-           while (len-- > 0) {
-               unsigned long aulong;
+            while (len-- > 0) {
+                unsigned long aulong;
                 SHIFT_VAR(utf8, s, strend, aulong, datumtype, needs_swap);
-               if (!checksum)
-                   mPUSHu(aulong);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)aulong;
-               else
-                   cuv += aulong;
-           }
-           break;
+                if (!checksum)
+                    mPUSHu(aulong);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)aulong;
+                else
+                    cuv += aulong;
+            }
+            break;
 #else
             /* FALLTHROUGH */
 #endif
-       case 'V':
-       case 'N':
-       case 'L':
-           while (len-- > 0) {
-               U32 au32;
+        case 'V':
+        case 'N':
+        case 'L':
+            while (len-- > 0) {
+                U32 au32;
 #if U32SIZE > SIZE32
-               au32 = 0;
+                au32 = 0;
 #endif
                 SHIFT32(utf8, s, strend, &au32, datumtype, needs_swap);
-               if (datumtype == 'N')
-                   au32 = PerlSock_ntohl(au32);
-               if (datumtype == 'V')
-                   au32 = vtohl(au32);
-               if (!checksum)
-                   mPUSHu(au32);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)au32;
-               else
-                   cuv += au32;
-           }
-           break;
-       case 'V' | TYPE_IS_SHRIEKING:
-       case 'N' | TYPE_IS_SHRIEKING:
-           while (len-- > 0) {
-               I32 ai32;
+                if (datumtype == 'N')
+                    au32 = PerlSock_ntohl(au32);
+                if (datumtype == 'V')
+                    au32 = vtohl(au32);
+                if (!checksum)
+                    mPUSHu(au32);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)au32;
+                else
+                    cuv += au32;
+            }
+            break;
+        case 'V' | TYPE_IS_SHRIEKING:
+        case 'N' | TYPE_IS_SHRIEKING:
+            while (len-- > 0) {
+                I32 ai32;
 #if U32SIZE > SIZE32
-               ai32 = 0;
+                ai32 = 0;
 #endif
                 SHIFT32(utf8, s, strend, &ai32, datumtype, needs_swap);
                 /* There should never be any byte swapping here.  */
                 assert(!TYPE_ENDIANNESS(datumtype));
-               if (datumtype == ('N' | TYPE_IS_SHRIEKING))
-                   ai32 = (I32)PerlSock_ntohl((U32)ai32);
-               if (datumtype == ('V' | TYPE_IS_SHRIEKING))
-                   ai32 = (I32)vtohl((U32)ai32);
-               if (!checksum)
-                   mPUSHi(ai32);
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)ai32;
-               else
-                   cuv += ai32;
-           }
-           break;
-       case 'p':
-           while (len-- > 0) {
-               const char *aptr;
+                if (datumtype == ('N' | TYPE_IS_SHRIEKING))
+                    ai32 = (I32)PerlSock_ntohl((U32)ai32);
+                if (datumtype == ('V' | TYPE_IS_SHRIEKING))
+                    ai32 = (I32)vtohl((U32)ai32);
+                if (!checksum)
+                    mPUSHi(ai32);
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)ai32;
+                else
+                    cuv += ai32;
+            }
+            break;
+        case 'p':
+            while (len-- > 0) {
+                const char *aptr;
                 SHIFT_VAR(utf8, s, strend, aptr, datumtype, needs_swap);
-               /* newSVpv generates undef if aptr is NULL */
-               mPUSHs(newSVpv(aptr, 0));
-           }
-           break;
-       case 'w':
-           {
-               UV auv = 0;
-               U32 bytes = 0;
-
-               while (len > 0 && s < strend) {
-                   U8 ch;
-                   ch = SHIFT_BYTE(utf8, s, strend, datumtype);
-                   auv = (auv << 7) | (ch & 0x7f);
-                   /* UTF8_IS_XXXXX not right here - using constant 0x80 */
-                   if (ch < 0x80) {
-                       bytes = 0;
-                       mPUSHu(auv);
-                       len--;
-                       auv = 0;
-                       continue;
-                   }
-                   if (++bytes >= sizeof(UV)) {        /* promote to string */
-                       const char *t;
-
-                       sv = Perl_newSVpvf(aTHX_ "%.*"UVuf, (int)TYPE_DIGITS(UV), auv);
-                       while (s < strend) {
-                           ch = SHIFT_BYTE(utf8, s, strend, datumtype);
-                           sv = mul128(sv, (U8)(ch & 0x7f));
-                           if (!(ch & 0x80)) {
-                               bytes = 0;
-                               break;
-                           }
-                       }
-                       t = SvPV_nolen_const(sv);
-                       while (*t == '0')
-                           t++;
-                       sv_chop(sv, t);
-                       mPUSHs(sv);
-                       len--;
-                       auv = 0;
-                   }
-               }
-               if ((s >= strend) && bytes)
-                   Perl_croak(aTHX_ "Unterminated compressed integer in unpack");
-           }
-           break;
-       case 'P':
-           if (symptr->howlen == e_star)
-               Perl_croak(aTHX_ "'P' must have an explicit size in unpack");
-           EXTEND(SP, 1);
-           if (s + sizeof(char*) <= strend) {
-               char *aptr;
+                /* newSVpv generates undef if aptr is NULL */
+                mPUSHs(newSVpv(aptr, 0));
+            }
+            break;
+        case 'w':
+            {
+                UV auv = 0;
+                size_t bytes = 0;
+
+                while (len > 0 && s < strend) {
+                    U8 ch;
+                    ch = SHIFT_BYTE(utf8, s, strend, datumtype);
+                    auv = (auv << 7) | (ch & 0x7f);
+                    /* UTF8_IS_XXXXX not right here because this is a BER, not
+                     * UTF-8 format - using constant 0x80 */
+                    if (ch < 0x80) {
+                        bytes = 0;
+                        mPUSHu(auv);
+                        len--;
+                        auv = 0;
+                        continue;
+                    }
+                    if (++bytes >= sizeof(UV)) {       /* promote to string */
+                        const char *t;
+
+                        sv = Perl_newSVpvf(aTHX_ "%.*" UVuf,
+                                                 (int)TYPE_DIGITS(UV), auv);
+                        while (s < strend) {
+                            ch = SHIFT_BYTE(utf8, s, strend, datumtype);
+                            sv = mul128(sv, (U8)(ch & 0x7f));
+                            if (!(ch & 0x80)) {
+                                bytes = 0;
+                                break;
+                            }
+                        }
+                        t = SvPV_nolen_const(sv);
+                        while (*t == '0')
+                            t++;
+                        sv_chop(sv, t);
+                        mPUSHs(sv);
+                        len--;
+                        auv = 0;
+                    }
+                }
+                if ((s >= strend) && bytes)
+                    Perl_croak(aTHX_ "Unterminated compressed integer in unpack");
+            }
+            break;
+        case 'P':
+            if (symptr->howlen == e_star)
+                Perl_croak(aTHX_ "'P' must have an explicit size in unpack");
+            EXTEND(SP, 1);
+            if (s + sizeof(char*) <= strend) {
+                char *aptr;
                 SHIFT_VAR(utf8, s, strend, aptr, datumtype, needs_swap);
-               /* newSVpvn generates undef if aptr is NULL */
-               PUSHs(newSVpvn_flags(aptr, len, SVs_TEMP));
-           }
-           break;
+                /* newSVpvn generates undef if aptr is NULL */
+                PUSHs(newSVpvn_flags(aptr, len, SVs_TEMP));
+            }
+            break;
 #if defined(HAS_QUAD) && IVSIZE >= 8
-       case 'q':
-           while (len-- > 0) {
-               Quad_t aquad;
+        case 'q':
+            while (len-- > 0) {
+                Quad_t aquad;
                 SHIFT_VAR(utf8, s, strend, aquad, datumtype, needs_swap);
-               if (!checksum)
+                if (!checksum)
                     mPUSHs(newSViv((IV)aquad));
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)aquad;
-               else
-                   cuv += aquad;
-           }
-           break;
-       case 'Q':
-           while (len-- > 0) {
-               Uquad_t auquad;
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)aquad;
+                else
+                    cuv += aquad;
+            }
+            break;
+        case 'Q':
+            while (len-- > 0) {
+                Uquad_t auquad;
                 SHIFT_VAR(utf8, s, strend, auquad, datumtype, needs_swap);
-               if (!checksum)
-                   mPUSHs(newSVuv((UV)auquad));
-               else if (checksum > bits_in_uv)
-                   cdouble += (NV)auquad;
-               else
-                   cuv += auquad;
-           }
-           break;
+                if (!checksum)
+                    mPUSHs(newSVuv((UV)auquad));
+                else if (checksum > bits_in_uv)
+                    cdouble += (NV)auquad;
+                else
+                    cuv += auquad;
+            }
+            break;
 #endif
-       /* float and double added gnb@melba.bby.oz.au 22/11/89 */
-       case 'f':
-           while (len-- > 0) {
-               float afloat;
+        /* float and double added gnb@melba.bby.oz.au 22/11/89 */
+        case 'f':
+            while (len-- > 0) {
+                float afloat;
                 SHIFT_VAR(utf8, s, strend, afloat, datumtype, needs_swap);
-               if (!checksum)
-                   mPUSHn(afloat);
-               else
-                   cdouble += afloat;
-           }
-           break;
-       case 'd':
-           while (len-- > 0) {
-               double adouble;
+                if (!checksum)
+                    mPUSHn(afloat);
+                else
+                    cdouble += afloat;
+            }
+            break;
+        case 'd':
+            while (len-- > 0) {
+                double adouble;
                 SHIFT_VAR(utf8, s, strend, adouble, datumtype, needs_swap);
-               if (!checksum)
-                   mPUSHn(adouble);
-               else
-                   cdouble += adouble;
-           }
-           break;
-       case 'F':
-           while (len-- > 0) {
-               NV_bytes anv;
+                if (!checksum)
+                    mPUSHn(adouble);
+                else
+                    cdouble += adouble;
+            }
+            break;
+        case 'F':
+            while (len-- > 0) {
+                NV_bytes anv;
                 SHIFT_BYTES(utf8, s, strend, anv.bytes, sizeof(anv.bytes),
                             datumtype, needs_swap);
-               if (!checksum)
-                   mPUSHn(anv.nv);
-               else
-                   cdouble += anv.nv;
-           }
-           break;
-#if defined(HAS_LONG_DOUBLE) && defined(USE_LONG_DOUBLE)
-       case 'D':
-           while (len-- > 0) {
-               ld_bytes aldouble;
+                if (!checksum)
+                    mPUSHn(anv.nv);
+                else
+                    cdouble += anv.nv;
+            }
+            break;
+#if defined(HAS_LONG_DOUBLE)
+        case 'D':
+            while (len-- > 0) {
+                ld_bytes aldouble;
                 SHIFT_BYTES(utf8, s, strend, aldouble.bytes,
                             sizeof(aldouble.bytes), datumtype, needs_swap);
                 /* The most common long double format, the x86 80-bit
@@ -1710,143 +1714,133 @@ S_unpack_rec(pTHX_ tempsym_t* symptr, const char *s, const char *strbeg, const c
                  * Note that trying to unpack 'long doubles' of 'long
                  * doubles' packed in another system is in the general
                  * case doomed without having more detail. */
-               if (!checksum)
-                   mPUSHn(aldouble.ld);
-               else
-                   cdouble += aldouble.ld;
-           }
-           break;
+                if (!checksum)
+                    mPUSHn(aldouble.ld);
+                else
+                    cdouble += aldouble.ld;
+            }
+            break;
 #endif
-       case 'u':
-           if (!checksum) {
+        case 'u':
+            if (!checksum) {
                 const STRLEN l = (STRLEN) (strend - s) * 3 / 4;
-               sv = sv_2mortal(newSV(l));
-               if (l) SvPOK_on(sv);
-           }
-           if (utf8) {
-               while (next_uni_uu(aTHX_ &s, strend, &len)) {
-                   I32 a, b, c, d;
-                   char hunk[3];
-
-                   while (len > 0) {
-                       next_uni_uu(aTHX_ &s, strend, &a);
-                       next_uni_uu(aTHX_ &s, strend, &b);
-                       next_uni_uu(aTHX_ &s, strend, &c);
-                       next_uni_uu(aTHX_ &s, strend, &d);
-                       hunk[0] = (char)((a << 2) | (b >> 4));
-                       hunk[1] = (char)((b << 4) | (c >> 2));
-                       hunk[2] = (char)((c << 6) | d);
-                       if (!checksum)
-                           sv_catpvn(sv, hunk, (len > 3) ? 3 : len);
-                       len -= 3;
-                   }
-                   if (s < strend) {
-                       if (*s == '\n') {
-                            s++;
-                        }
-                       else {
-                           /* possible checksum byte */
-                           const char *skip = s+UTF8SKIP(s);
-                           if (skip < strend && *skip == '\n')
-                                s = skip+1;
-                       }
-                   }
-               }
-           } else {
-               while (s < strend && *s > ' ' && ISUUCHAR(*s)) {
-                   I32 a, b, c, d;
-                   char hunk[3];
-
-                   len = PL_uudmap[*(U8*)s++] & 077;
-                   while (len > 0) {
-                       if (s < strend && ISUUCHAR(*s))
-                           a = PL_uudmap[*(U8*)s++] & 077;
-                       else
-                           a = 0;
-                       if (s < strend && ISUUCHAR(*s))
-                           b = PL_uudmap[*(U8*)s++] & 077;
-                       else
-                           b = 0;
-                       if (s < strend && ISUUCHAR(*s))
-                           c = PL_uudmap[*(U8*)s++] & 077;
-                       else
-                           c = 0;
-                       if (s < strend && ISUUCHAR(*s))
-                           d = PL_uudmap[*(U8*)s++] & 077;
-                       else
-                           d = 0;
-                       hunk[0] = (char)((a << 2) | (b >> 4));
-                       hunk[1] = (char)((b << 4) | (c >> 2));
-                       hunk[2] = (char)((c << 6) | d);
-                       if (!checksum)
-                           sv_catpvn(sv, hunk, (len > 3) ? 3 : len);
-                       len -= 3;
-                   }
-                   if (*s == '\n')
-                       s++;
-                   else        /* possible checksum byte */
-                       if (s + 1 < strend && s[1] == '\n')
-                           s += 2;
-               }
-           }
-           if (!checksum)
-               XPUSHs(sv);
-           break;
-       }
-
-       if (checksum) {
-           if (strchr("fFdD", TYPE_NO_MODIFIERS(datumtype)) ||
-             (checksum > bits_in_uv &&
-              strchr("cCsSiIlLnNUWvVqQjJ", TYPE_NO_MODIFIERS(datumtype))) ) {
-               NV trouble, anv;
+                sv = sv_2mortal(newSV(l));
+                if (l) {
+                    SvPOK_on(sv);
+                    *SvEND(sv) = '\0';
+                }
+            }
+
+            /* Note that all legal uuencoded strings are ASCII printables, so
+             * have the same representation under UTF-8 vs not.  This means we
+             * can ignore UTF8ness on legal input.  For illegal we stop at the
+             * first failure, and don't report where/what that is, so again we
+             * can ignore UTF8ness */
+
+            while (s < strend && *s != ' ' && ISUUCHAR(*s)) {
+                I32 a, b, c, d;
+                char hunk[3];
+
+                len = PL_uudmap[*(U8*)s++] & 077;
+                while (len > 0) {
+                    if (s < strend && ISUUCHAR(*s))
+                        a = PL_uudmap[*(U8*)s++] & 077;
+                    else
+                        a = 0;
+                    if (s < strend && ISUUCHAR(*s))
+                        b = PL_uudmap[*(U8*)s++] & 077;
+                    else
+                        b = 0;
+                    if (s < strend && ISUUCHAR(*s))
+                        c = PL_uudmap[*(U8*)s++] & 077;
+                    else
+                        c = 0;
+                    if (s < strend && ISUUCHAR(*s))
+                        d = PL_uudmap[*(U8*)s++] & 077;
+                    else
+                        d = 0;
+                    hunk[0] = (char)((a << 2) | (b >> 4));
+                    hunk[1] = (char)((b << 4) | (c >> 2));
+                    hunk[2] = (char)((c << 6) | d);
+                    if (!checksum)
+                        sv_catpvn(sv, hunk, (len > 3) ? 3 : len);
+                    len -= 3;
+                }
+                if (*s == '\n')
+                    s++;
+                else   /* possible checksum byte */
+                    if (s + 1 < strend && s[1] == '\n')
+                        s += 2;
+            }
+            if (!checksum)
+                XPUSHs(sv);
+            break;
+        } /* End of switch */
+
+        if (checksum) {
+            if (memCHRs("fFdD", TYPE_NO_MODIFIERS(datumtype)) ||
+              (checksum > bits_in_uv &&
+               memCHRs("cCsSiIlLnNUWvVqQjJ", TYPE_NO_MODIFIERS(datumtype))) ) {
+                NV trouble, anv;
 
                 anv = (NV) (1 << (checksum & 15));
-               while (checksum >= 16) {
-                   checksum -= 16;
-                   anv *= 65536.0;
-               }
-               while (cdouble < 0.0)
-                   cdouble += anv;
-               cdouble = Perl_modf(cdouble / anv, &trouble) * anv;
-               sv = newSVnv(cdouble);
-           }
-           else {
-               if (checksum < bits_in_uv) {
-                   UV mask = ((UV)1 << checksum) - 1;
-                   cuv &= mask;
-               }
-               sv = newSVuv(cuv);
-           }
-           mXPUSHs(sv);
-           checksum = 0;
-       }
+                while (checksum >= 16) {
+                    checksum -= 16;
+                    anv *= 65536.0;
+                }
+                while (cdouble < 0.0)
+                    cdouble += anv;
+                cdouble = Perl_modf(cdouble / anv, &trouble);
+#ifdef LONGDOUBLE_DOUBLEDOUBLE
+                /* Workaround for powerpc doubledouble modfl bug:
+                 * close to 1.0L and -1.0L cdouble is 0, and trouble
+                 * is cdouble / anv. */
+                if (trouble != Perl_ceil(trouble)) {
+                  cdouble = trouble;
+                  if (cdouble >  1.0L) cdouble -= 1.0L;
+                  if (cdouble < -1.0L) cdouble += 1.0L;
+                }
+#endif
+                cdouble *= anv;
+                sv = newSVnv(cdouble);
+            }
+            else {
+                if (checksum < bits_in_uv) {
+                    UV mask = nBIT_MASK(checksum);
+                    cuv &= mask;
+                }
+                sv = newSVuv(cuv);
+            }
+            mXPUSHs(sv);
+            checksum = 0;
+        }
 
         if (symptr->flags & FLAG_SLASH){
             if (SP - PL_stack_base - start_sp_offset <= 0)
-               break;
+                break;
             if( next_symbol(symptr) ){
               if( symptr->howlen == e_number )
-               Perl_croak(aTHX_ "Count after length/code in unpack" );
+                Perl_croak(aTHX_ "Count after length/code in unpack" );
               if( beyond ){
-               /* ...end of char buffer then no decent length available */
-               Perl_croak(aTHX_ "length/code after end of string in unpack" );
+                /* ...end of char buffer then no decent length available */
+                Perl_croak(aTHX_ "length/code after end of string in unpack" );
               } else {
-               /* take top of stack (hope it's numeric) */
+                /* take top of stack (hope it's numeric) */
                 len = POPi;
                 if( len < 0 )
                     Perl_croak(aTHX_ "Negative '/' count in unpack" );
               }
             } else {
-               Perl_croak(aTHX_ "Code missing after '/' in unpack" );
+                Perl_croak(aTHX_ "Code missing after '/' in unpack" );
             }
             datumtype = symptr->code;
             explicit_length = FALSE;
-           goto redo_switch;
+            goto redo_switch;
         }
     }
 
     if (new_s)
-       *new_s = s;
+        *new_s = s;
     PUTBACK;
     return SP - PL_stack_base - start_sp_offset;
 }
@@ -1855,19 +1849,19 @@ PP(pp_unpack)
 {
     dSP;
     dPOPPOPssrl;
-    I32 gimme = GIMME_V;
+    U8 gimme = GIMME_V;
     STRLEN llen;
     STRLEN rlen;
     const char *pat = SvPV_const(left,  llen);
     const char *s   = SvPV_const(right, rlen);
     const char *strend = s + rlen;
     const char *patend = pat + llen;
-    I32 cnt;
+    SSize_t cnt;
 
     PUTBACK;
     cnt = unpackstring(pat, patend, s, strend,
-                    ((gimme == G_SCALAR) ? FLAG_UNPACK_ONLY_ONE : 0)
-                    | (DO_UTF8(right) ? FLAG_DO_UTF8 : 0));
+                     ((gimme == G_SCALAR) ? FLAG_UNPACK_ONLY_ONE : 0)
+                     | (DO_UTF8(right) ? FLAG_DO_UTF8 : 0));
 
     SPAGAIN;
     if ( !cnt && gimme == G_SCALAR )
@@ -1876,23 +1870,23 @@ PP(pp_unpack)
 }
 
 STATIC U8 *
-doencodes(U8 *h, const char *s, I32 len)
+doencodes(U8 *h, const U8 *s, SSize_t len)
 {
     *h++ = PL_uuemap[len];
     while (len > 2) {
-       *h++ = PL_uuemap[(077 & (s[0] >> 2))];
-       *h++ = PL_uuemap[(077 & (((s[0] << 4) & 060) | ((s[1] >> 4) & 017)))];
-       *h++ = PL_uuemap[(077 & (((s[1] << 2) & 074) | ((s[2] >> 6) & 03)))];
-       *h++ = PL_uuemap[(077 & (s[2] & 077))];
-       s += 3;
-       len -= 3;
+        *h++ = PL_uuemap[(077 & (s[0] >> 2))];
+        *h++ = PL_uuemap[(077 & (((s[0] << 4) & 060) | ((s[1] >> 4) & 017)))];
+        *h++ = PL_uuemap[(077 & (((s[1] << 2) & 074) | ((s[2] >> 6) & 03)))];
+        *h++ = PL_uuemap[(077 & (s[2] & 077))];
+        s += 3;
+        len -= 3;
     }
     if (len > 0) {
-        const char r = (len > 1 ? s[1] : '\0');
-       *h++ = PL_uuemap[(077 & (s[0] >> 2))];
-       *h++ = PL_uuemap[(077 & (((s[0] << 4) & 060) | ((r >> 4) & 017)))];
-       *h++ = PL_uuemap[(077 & ((r << 2) & 074))];
-       *h++ = PL_uuemap[0];
+        const U8 r = (len > 1 ? s[1] : '\0');
+        *h++ = PL_uuemap[(077 & (s[0] >> 2))];
+        *h++ = PL_uuemap[(077 & (((s[0] << 4) & 060) | ((r >> 4) & 017)))];
+        *h++ = PL_uuemap[(077 & ((r << 2) & 074))];
+        *h++ = PL_uuemap[0];
     }
     *h++ = '\n';
     return h;
@@ -1915,8 +1909,8 @@ S_is_an_int(pTHX_ const char *s, STRLEN l)
       break;
     case '+':
       if (!skip) {
-       SvREFCNT_dec(result);
-       return (NULL);
+        SvREFCNT_dec(result);
+        return (NULL);
       }
       break;
     case '0':
@@ -1931,7 +1925,7 @@ S_is_an_int(pTHX_ const char *s, STRLEN l)
     case '9':
       skip = 0;
       if (!ignore) {
-       *(out++) = *s;
+        *(out++) = *s;
       }
       break;
     case '.':
@@ -1961,13 +1955,13 @@ S_div128(pTHX_ SV *pnum, bool *done)
 
     *done = 1;
     while (*t) {
-       const int i = m * 10 + (*t - '0');
-       const int r = (i >> 7); /* r < 10 */
-       m = i & 0x7F;
-       if (r) {
-           *done = 0;
-       }
-       *(t++) = '0' + r;
+        const int i = m * 10 + (*t - '0');
+        const int r = (i >> 7); /* r < 10 */
+        m = i & 0x7F;
+        if (r) {
+            *done = 0;
+        }
+        *(t++) = '0' + r;
     }
     *(t++) = '\0';
     SvCUR_set(pnum, (STRLEN) (t - s));
@@ -1977,7 +1971,7 @@ S_div128(pTHX_ SV *pnum, bool *done)
 /*
 =for apidoc packlist
 
-The engine implementing pack() Perl function.
+The engine implementing C<pack()> Perl function.
 
 =cut
 */
@@ -1995,7 +1989,7 @@ Perl_packlist(pTHX_ SV *cat, const char *pat, const char *patend, SV **beglist,
        Also make sure any UTF8 flag is loaded */
     SvPV_force_nolen(cat);
     if (DO_UTF8(cat))
-       sym.flags |= FLAG_PARSE_UTF8 | FLAG_DO_UTF8;
+        sym.flags |= FLAG_PARSE_UTF8 | FLAG_DO_UTF8;
 
     (void)pack_rec( cat, &sym, beglist, endlist );
 }
@@ -2013,11 +2007,11 @@ marked_upgrade(pTHX_ SV *sv, tempsym_t *sym_ptr) {
     from_start = SvPVX_const(sv);
     from_end = from_start + SvCUR(sv);
     for (from_ptr = from_start; from_ptr < from_end; from_ptr++)
-       if (!NATIVE_BYTE_IS_INVARIANT(*from_ptr)) break;
+        if (!NATIVE_BYTE_IS_INVARIANT(*from_ptr)) break;
     if (from_ptr == from_end) {
-       /* Simple case: no character needs to be changed */
-       SvUTF8_on(sv);
-       return;
+        /* Simple case: no character needs to be changed */
+        SvUTF8_on(sv);
+        return;
     }
 
     len = (from_end-from_ptr)*UTF8_EXPAND+(from_ptr-from_start)+1;
@@ -2027,38 +2021,38 @@ marked_upgrade(pTHX_ SV *sv, tempsym_t *sym_ptr) {
 
     Newx(marks, sym_ptr->level+2, const char *);
     for (group=sym_ptr; group; group = group->previous)
-       marks[group->level] = from_start + group->strbeg;
+        marks[group->level] = from_start + group->strbeg;
     marks[sym_ptr->level+1] = from_end+1;
     for (m = marks; *m < from_ptr; m++)
-       *m = to_start + (*m-from_start);
+        *m = to_start + (*m-from_start);
 
     for (;from_ptr < from_end; from_ptr++) {
-       while (*m == from_ptr) *m++ = to_ptr;
-       to_ptr = (char *) uvchr_to_utf8((U8 *) to_ptr, *(U8 *) from_ptr);
+        while (*m == from_ptr) *m++ = to_ptr;
+        to_ptr = (char *) uvchr_to_utf8((U8 *) to_ptr, *(U8 *) from_ptr);
     }
     *to_ptr = 0;
 
     while (*m == from_ptr) *m++ = to_ptr;
     if (m != marks + sym_ptr->level+1) {
-       Safefree(marks);
-       Safefree(to_start);
-       Perl_croak(aTHX_ "panic: marks beyond string end, m=%p, marks=%p, "
-                  "level=%d", m, marks, sym_ptr->level);
+        Safefree(marks);
+        Safefree(to_start);
+        Perl_croak(aTHX_ "panic: marks beyond string end, m=%p, marks=%p, "
+                   "level=%d", m, marks, sym_ptr->level);
     }
     for (group=sym_ptr; group; group = group->previous)
-       group->strbeg = marks[group->level] - to_start;
+        group->strbeg = marks[group->level] - to_start;
     Safefree(marks);
 
     if (SvOOK(sv)) {
-       if (SvIVX(sv)) {
-           SvLEN_set(sv, SvLEN(sv) + SvIVX(sv));
-           from_start -= SvIVX(sv);
-           SvIV_set(sv, 0);
-       }
-       SvFLAGS(sv) &= ~SVf_OOK;
+        if (SvIVX(sv)) {
+            SvLEN_set(sv, SvLEN(sv) + SvIVX(sv));
+            from_start -= SvIVX(sv);
+            SvIV_set(sv, 0);
+        }
+        SvFLAGS(sv) &= ~SVf_OOK;
     }
     if (SvLEN(sv) != 0)
-       Safefree(from_start);
+        Safefree(from_start);
     SvPV_set(sv, to_start);
     SvCUR_set(sv, to_ptr - to_start);
     SvLEN_set(sv, len);
@@ -2082,12 +2076,34 @@ S_sv_exp_grow(pTHX_ SV *sv, STRLEN needed) {
     return SvGROW(sv, len+extend+1);
 }
 
+static SV *
+S_sv_check_infnan(pTHX_ SV *sv, I32 datumtype)
+{
+    SvGETMAGIC(sv);
+    if (UNLIKELY(SvAMAGIC(sv)))
+        sv = sv_2num(sv);
+    if (UNLIKELY(isinfnansv(sv))) {
+        const I32 c = TYPE_NO_MODIFIERS(datumtype);
+        const NV nv = SvNV_nomg(sv);
+        if (c == 'w')
+            Perl_croak(aTHX_ "Cannot compress %" NVgf " in pack", nv);
+        else
+            Perl_croak(aTHX_ "Cannot pack %" NVgf " with '%c'", nv, (int) c);
+    }
+    return sv;
+}
+
+#define SvIV_no_inf(sv,d) \
+        ((sv) = S_sv_check_infnan(aTHX_ sv,d), SvIV_nomg(sv))
+#define SvUV_no_inf(sv,d) \
+        ((sv) = S_sv_check_infnan(aTHX_ sv,d), SvUV_nomg(sv))
+
 STATIC
 SV **
 S_pack_rec(pTHX_ SV *cat, tempsym_t* symptr, SV **beglist, SV **endlist )
 {
     tempsym_t lookahead;
-    I32 items  = endlist - beglist;
+    SSize_t items  = endlist - beglist;
     bool found = next_symbol(symptr);
     bool utf8 = (symptr->flags & FLAG_PARSE_UTF8) ? 1 : 0;
     bool warn_utf8 = ckWARN(WARN_UTF8);
@@ -2096,1022 +2112,1021 @@ S_pack_rec(pTHX_ SV *cat, tempsym_t* symptr, SV **beglist, SV **endlist )
     PERL_ARGS_ASSERT_PACK_REC;
 
     if (symptr->level == 0 && found && symptr->code == 'U') {
-       marked_upgrade(aTHX_ cat, symptr);
-       symptr->flags |= FLAG_DO_UTF8;
-       utf8 = 0;
+        marked_upgrade(aTHX_ cat, symptr);
+        symptr->flags |= FLAG_DO_UTF8;
+        utf8 = 0;
     }
     symptr->strbeg = SvCUR(cat);
 
     while (found) {
-       SV *fromstr;
-       STRLEN fromlen;
-       I32 len;
-       SV *lengthcode = NULL;
+        SV *fromstr;
+        STRLEN fromlen;
+        SSize_t len;
+        SV *lengthcode = NULL;
         I32 datumtype = symptr->code;
         howlen_t howlen = symptr->howlen;
-       char *start = SvPVX(cat);
-       char *cur   = start + SvCUR(cat);
+        char *start = SvPVX(cat);
+        char *cur   = start + SvCUR(cat);
         bool needs_swap;
 
-#define NEXTFROM (lengthcode ? lengthcode : items-- > 0 ? *beglist++ : &PL_sv_no)
+#define NEXTFROM (lengthcode ? lengthcode : items > 0 ? (--items, *beglist++) : &PL_sv_no)
+#define PEEKFROM (lengthcode ? lengthcode : items > 0 ? *beglist : &PL_sv_no)
 
         switch (howlen) {
-         case e_star:
-           len = strchr("@Xxu", TYPE_NO_MODIFIERS(datumtype)) ?
-               0 : items;
-           break;
-         default:
-           /* e_no_len and e_number */
-           len = symptr->length;
-           break;
+          case e_star:
+            len = memCHRs("@Xxu", TYPE_NO_MODIFIERS(datumtype)) ?
+                0 : items;
+            break;
+          default:
+            /* e_no_len and e_number */
+            len = symptr->length;
+            break;
         }
 
-       if (len) {
-           packprops_t props = packprops[TYPE_NO_ENDIANNESS(datumtype)];
+        if (len) {
+            packprops_t props = packprops[TYPE_NO_ENDIANNESS(datumtype)];
 
-           if (props && !(props & PACK_SIZE_UNPREDICTABLE)) {
-               /* We can process this letter. */
-               STRLEN size = props & PACK_SIZE_MASK;
-               GROWING(utf8, cat, start, cur, (STRLEN) len * size);
-           }
+            if (props && !(props & PACK_SIZE_UNPREDICTABLE)) {
+                /* We can process this letter. */
+                STRLEN size = props & PACK_SIZE_MASK;
+                GROWING2(utf8, cat, start, cur, size, (STRLEN)len);
+            }
         }
 
         /* Look ahead for next symbol. Do we have code/code? */
         lookahead = *symptr;
         found = next_symbol(&lookahead);
-       if (symptr->flags & FLAG_SLASH) {
-           IV count;
-           if (!found) Perl_croak(aTHX_ "Code missing after '/' in pack");
-           if (strchr("aAZ", lookahead.code)) {
-               if (lookahead.howlen == e_number) count = lookahead.length;
-               else {
-                   if (items > 0) {
-                       count = sv_len_utf8(*beglist);
-                   }
-                   else count = 0;
-                   if (lookahead.code == 'Z') count++;
-               }
-           } else {
-               if (lookahead.howlen == e_number && lookahead.length < items)
-                   count = lookahead.length;
-               else count = items;
-           }
-           lookahead.howlen = e_number;
-           lookahead.length = count;
-           lengthcode = sv_2mortal(newSViv(count));
-       }
+        if (symptr->flags & FLAG_SLASH) {
+            IV count;
+            if (!found) Perl_croak(aTHX_ "Code missing after '/' in pack");
+            if (memCHRs("aAZ", lookahead.code)) {
+                if (lookahead.howlen == e_number) count = lookahead.length;
+                else {
+                    if (items > 0) {
+                        count = sv_len_utf8(*beglist);
+                    }
+                    else count = 0;
+                    if (lookahead.code == 'Z') count++;
+                }
+            } else {
+                if (lookahead.howlen == e_number && lookahead.length < items)
+                    count = lookahead.length;
+                else count = items;
+            }
+            lookahead.howlen = e_number;
+            lookahead.length = count;
+            lengthcode = sv_2mortal(newSViv(count));
+        }
 
         needs_swap = NEEDS_SWAP(datumtype);
 
-       /* Code inside the switch must take care to properly update
-          cat (CUR length and '\0' termination) if it updated *cur and
-          doesn't simply leave using break */
-       switch(TYPE_NO_ENDIANNESS(datumtype)) {
-       default:
-           Perl_croak(aTHX_ "Invalid type '%c' in pack",
-                      (int) TYPE_NO_MODIFIERS(datumtype));
-       case '%':
-           Perl_croak(aTHX_ "'%%' may not be used in pack");
-
-       case '.' | TYPE_IS_SHRIEKING:
-       case '.':
-           if (howlen == e_star) from = start;
-           else if (len == 0) from = cur;
-           else {
-               tempsym_t *group = symptr;
-
-               while (--len && group) group = group->previous;
-               from = group ? start + group->strbeg : start;
-           }
-           fromstr = NEXTFROM;
-           len = SvIV(fromstr);
-           goto resize;
-       case '@' | TYPE_IS_SHRIEKING:
-       case '@':
-           from = start + symptr->strbeg;
-         resize:
-           if (utf8  && !(datumtype & TYPE_IS_SHRIEKING))
-               if (len >= 0) {
-                   while (len && from < cur) {
-                       from += UTF8SKIP(from);
-                       len--;
-                   }
-                   if (from > cur)
-                       Perl_croak(aTHX_ "Malformed UTF-8 string in pack");
-                   if (len) {
-                       /* Here we know from == cur */
-                     grow:
-                       GROWING(0, cat, start, cur, len);
-                       Zero(cur, len, char);
-                       cur += len;
-                   } else if (from < cur) {
-                       len = cur - from;
-                       goto shrink;
-                   } else goto no_change;
-               } else {
-                   cur = from;
-                   len = -len;
-                   goto utf8_shrink;
-               }
-           else {
-               len -= cur - from;
-               if (len > 0) goto grow;
-               if (len == 0) goto no_change;
-               len = -len;
-               goto shrink;
-           }
-           break;
-
-       case '(': {
+        /* Code inside the switch must take care to properly update
+           cat (CUR length and '\0' termination) if it updated *cur and
+           doesn't simply leave using break */
+        switch (TYPE_NO_ENDIANNESS(datumtype)) {
+        default:
+            Perl_croak(aTHX_ "Invalid type '%c' in pack",
+                       (int) TYPE_NO_MODIFIERS(datumtype));
+        case '%':
+            Perl_croak(aTHX_ "'%%' may not be used in pack");
+
+        case '.' | TYPE_IS_SHRIEKING:
+        case '.':
+            if (howlen == e_star) from = start;
+            else if (len == 0) from = cur;
+            else {
+                tempsym_t *group = symptr;
+
+                while (--len && group) group = group->previous;
+                from = group ? start + group->strbeg : start;
+            }
+            fromstr = NEXTFROM;
+            len = SvIV_no_inf(fromstr, datumtype);
+            goto resize;
+        case '@' | TYPE_IS_SHRIEKING:
+        case '@':
+            from = start + symptr->strbeg;
+          resize:
+            if (utf8  && !(datumtype & TYPE_IS_SHRIEKING))
+                if (len >= 0) {
+                    while (len && from < cur) {
+                        from += UTF8SKIP(from);
+                        len--;
+                    }
+                    if (from > cur)
+                        Perl_croak(aTHX_ "Malformed UTF-8 string in pack");
+                    if (len) {
+                        /* Here we know from == cur */
+                      grow:
+                        GROWING(0, cat, start, cur, len);
+                        Zero(cur, len, char);
+                        cur += len;
+                    } else if (from < cur) {
+                        len = cur - from;
+                        goto shrink;
+                    } else goto no_change;
+                } else {
+                    cur = from;
+                    len = -len;
+                    goto utf8_shrink;
+                }
+            else {
+                len -= cur - from;
+                if (len > 0) goto grow;
+                if (len == 0) goto no_change;
+                len = -len;
+                goto shrink;
+            }
+            break;
+
+        case '(': {
             tempsym_t savsym = *symptr;
-           U32 group_modifiers = TYPE_MODIFIERS(datumtype & ~symptr->flags);
-           symptr->flags |= group_modifiers;
+            U32 group_modifiers = TYPE_MODIFIERS(datumtype & ~symptr->flags);
+            symptr->flags |= group_modifiers;
             symptr->patend = savsym.grpend;
             symptr->level++;
-           symptr->previous = &lookahead;
-           while (len--) {
-               U32 was_utf8;
-               if (utf8) symptr->flags |=  FLAG_PARSE_UTF8;
-               else      symptr->flags &= ~FLAG_PARSE_UTF8;
-               was_utf8 = SvUTF8(cat);
-               symptr->patptr = savsym.grpbeg;
-               beglist = pack_rec(cat, symptr, beglist, endlist);
-               if (SvUTF8(cat) != was_utf8)
-                   /* This had better be an upgrade while in utf8==0 mode */
-                   utf8 = 1;
-
-               if (savsym.howlen == e_star && beglist == endlist)
-                   break;              /* No way to continue */
-           }
-           items = endlist - beglist;
-           lookahead.flags  = symptr->flags & ~group_modifiers;
-           goto no_change;
-       }
-       case 'X' | TYPE_IS_SHRIEKING:
-           if (!len)                   /* Avoid division by 0 */
-               len = 1;
-           if (utf8) {
-               char *hop, *last;
-               I32 l = len;
-               hop = last = start;
-               while (hop < cur) {
-                   hop += UTF8SKIP(hop);
-                   if (--l == 0) {
-                       last = hop;
-                       l = len;
-                   }
-               }
-               if (last > cur)
-                   Perl_croak(aTHX_ "Malformed UTF-8 string in pack");
-               cur = last;
-               break;
-           }
-           len = (cur-start) % len;
-           /* FALLTHROUGH */
-       case 'X':
-           if (utf8) {
-               if (len < 1) goto no_change;
-             utf8_shrink:
-               while (len > 0) {
-                   if (cur <= start)
-                       Perl_croak(aTHX_ "'%c' outside of string in pack",
-                                  (int) TYPE_NO_MODIFIERS(datumtype));
-                   while (--cur, UTF8_IS_CONTINUATION(*cur)) {
-                       if (cur <= start)
-                           Perl_croak(aTHX_ "'%c' outside of string in pack",
-                                      (int) TYPE_NO_MODIFIERS(datumtype));
-                   }
-                   len--;
-               }
-           } else {
-             shrink:
-               if (cur - start < len)
-                   Perl_croak(aTHX_ "'%c' outside of string in pack",
-                              (int) TYPE_NO_MODIFIERS(datumtype));
-               cur -= len;
-           }
-           if (cur < start+symptr->strbeg) {
-               /* Make sure group starts don't point into the void */
-               tempsym_t *group;
-               const STRLEN length = cur-start;
-               for (group = symptr;
-                    group && length < group->strbeg;
-                    group = group->previous) group->strbeg = length;
-               lookahead.strbeg = length;
-           }
-           break;
-       case 'x' | TYPE_IS_SHRIEKING: {
-           I32 ai32;
-           if (!len)                   /* Avoid division by 0 */
-               len = 1;
-           if (utf8) ai32 = utf8_length((U8 *) start, (U8 *) cur) % len;
-           else      ai32 = (cur - start) % len;
-           if (ai32 == 0) goto no_change;
-           len -= ai32;
-       }
-       /* FALLTHROUGH */
-       case 'x':
-           goto grow;
-       case 'A':
-       case 'Z':
-       case 'a': {
-           const char *aptr;
-
-           fromstr = NEXTFROM;
-           aptr = SvPV_const(fromstr, fromlen);
-           if (DO_UTF8(fromstr)) {
+            symptr->previous = &lookahead;
+            while (len--) {
+                U32 was_utf8;
+                if (utf8) symptr->flags |=  FLAG_PARSE_UTF8;
+                else      symptr->flags &= ~FLAG_PARSE_UTF8;
+                was_utf8 = SvUTF8(cat);
+                symptr->patptr = savsym.grpbeg;
+                beglist = pack_rec(cat, symptr, beglist, endlist);
+                if (SvUTF8(cat) != was_utf8)
+                    /* This had better be an upgrade while in utf8==0 mode */
+                    utf8 = 1;
+
+                if (savsym.howlen == e_star && beglist == endlist)
+                    break;             /* No way to continue */
+            }
+            items = endlist - beglist;
+            lookahead.flags  = symptr->flags & ~group_modifiers;
+            goto no_change;
+        }
+        case 'X' | TYPE_IS_SHRIEKING:
+            if (!len)                  /* Avoid division by 0 */
+                len = 1;
+            if (utf8) {
+                char *hop, *last;
+                SSize_t l = len;
+                hop = last = start;
+                while (hop < cur) {
+                    hop += UTF8SKIP(hop);
+                    if (--l == 0) {
+                        last = hop;
+                        l = len;
+                    }
+                }
+                if (last > cur)
+                    Perl_croak(aTHX_ "Malformed UTF-8 string in pack");
+                cur = last;
+                break;
+            }
+            len = (cur-start) % len;
+            /* FALLTHROUGH */
+        case 'X':
+            if (utf8) {
+                if (len < 1) goto no_change;
+              utf8_shrink:
+                while (len > 0) {
+                    if (cur <= start)
+                        Perl_croak(aTHX_ "'%c' outside of string in pack",
+                                   (int) TYPE_NO_MODIFIERS(datumtype));
+                    while (--cur, UTF8_IS_CONTINUATION(*cur)) {
+                        if (cur <= start)
+                            Perl_croak(aTHX_ "'%c' outside of string in pack",
+                                       (int) TYPE_NO_MODIFIERS(datumtype));
+                    }
+                    len--;
+                }
+            } else {
+              shrink:
+                if (cur - start < len)
+                    Perl_croak(aTHX_ "'%c' outside of string in pack",
+                               (int) TYPE_NO_MODIFIERS(datumtype));
+                cur -= len;
+            }
+            if (cur < start+symptr->strbeg) {
+                /* Make sure group starts don't point into the void */
+                tempsym_t *group;
+                const STRLEN length = cur-start;
+                for (group = symptr;
+                     group && length < group->strbeg;
+                     group = group->previous) group->strbeg = length;
+                lookahead.strbeg = length;
+            }
+            break;
+        case 'x' | TYPE_IS_SHRIEKING: {
+            SSize_t ai32;
+            if (!len)                  /* Avoid division by 0 */
+                len = 1;
+            if (utf8) ai32 = utf8_length((U8 *) start, (U8 *) cur) % len;
+            else      ai32 = (cur - start) % len;
+            if (ai32 == 0) goto no_change;
+            len -= ai32;
+        }
+        /* FALLTHROUGH */
+        case 'x':
+            goto grow;
+        case 'A':
+        case 'Z':
+        case 'a': {
+            const char *aptr;
+
+            fromstr = NEXTFROM;
+            aptr = SvPV_const(fromstr, fromlen);
+            if (DO_UTF8(fromstr)) {
                 const char *end, *s;
 
-               if (!utf8 && !SvUTF8(cat)) {
-                   marked_upgrade(aTHX_ cat, symptr);
-                   lookahead.flags |= FLAG_DO_UTF8;
-                   lookahead.strbeg = symptr->strbeg;
-                   utf8 = 1;
-                   start = SvPVX(cat);
-                   cur = start + SvCUR(cat);
-               }
-               if (howlen == e_star) {
-                   if (utf8) goto string_copy;
-                   len = fromlen+1;
-               }
-               s = aptr;
-               end = aptr + fromlen;
-               fromlen = datumtype == 'Z' ? len-1 : len;
-               while ((I32) fromlen > 0 && s < end) {
-                   s += UTF8SKIP(s);
-                   fromlen--;
-               }
-               if (s > end)
-                   Perl_croak(aTHX_ "Malformed UTF-8 string in pack");
-               if (utf8) {
-                   len = fromlen;
-                   if (datumtype == 'Z') len++;
-                   fromlen = s-aptr;
-                   len += fromlen;
-
-                   goto string_copy;
-               }
-               fromlen = len - fromlen;
-               if (datumtype == 'Z') fromlen--;
-               if (howlen == e_star) {
-                   len = fromlen;
-                   if (datumtype == 'Z') len++;
-               }
-               GROWING(0, cat, start, cur, len);
-               if (!uni_to_bytes(aTHX_ &aptr, end, cur, fromlen,
-                                 datumtype | TYPE_IS_PACK))
-                   Perl_croak(aTHX_ "panic: predicted utf8 length not available, "
-                              "for '%c', aptr=%p end=%p cur=%p, fromlen=%"UVuf,
-                              (int)datumtype, aptr, end, cur, (UV)fromlen);
-               cur += fromlen;
-               len -= fromlen;
-           } else if (utf8) {
-               if (howlen == e_star) {
-                   len = fromlen;
-                   if (datumtype == 'Z') len++;
-               }
-               if (len <= (I32) fromlen) {
-                   fromlen = len;
-                   if (datumtype == 'Z' && fromlen > 0) fromlen--;
-               }
-               /* assumes a byte expands to at most UTF8_EXPAND bytes on
-                  upgrade, so:
-                  expected_length <= from_len*UTF8_EXPAND + (len-from_len) */
-               GROWING(0, cat, start, cur, fromlen*(UTF8_EXPAND-1)+len);
-               len -= fromlen;
-               while (fromlen > 0) {
-                   cur = (char *) uvchr_to_utf8((U8 *) cur, * (U8 *) aptr);
-                   aptr++;
-                   fromlen--;
-               }
-           } else {
-             string_copy:
-               if (howlen == e_star) {
-                   len = fromlen;
-                   if (datumtype == 'Z') len++;
-               }
-               if (len <= (I32) fromlen) {
-                   fromlen = len;
-                   if (datumtype == 'Z' && fromlen > 0) fromlen--;
-               }
-               GROWING(0, cat, start, cur, len);
-               Copy(aptr, cur, fromlen, char);
-               cur += fromlen;
-               len -= fromlen;
-           }
-           memset(cur, datumtype == 'A' ? ' ' : '\0', len);
-           cur += len;
-           SvTAINT(cat);
-           break;
-       }
-       case 'B':
-       case 'b': {
-           const char *str, *end;
-           I32 l, field_len;
-           U8 bits;
-           bool utf8_source;
-           U32 utf8_flags;
-
-           fromstr = NEXTFROM;
-           str = SvPV_const(fromstr, fromlen);
-           end = str + fromlen;
-           if (DO_UTF8(fromstr)) {
-               utf8_source = TRUE;
-               utf8_flags  = warn_utf8 ? 0 : UTF8_ALLOW_ANY;
-           } else {
-               utf8_source = FALSE;
-               utf8_flags  = 0; /* Unused, but keep compilers happy */
-           }
-           if (howlen == e_star) len = fromlen;
-           field_len = (len+7)/8;
-           GROWING(utf8, cat, start, cur, field_len);
-           if (len > (I32)fromlen) len = fromlen;
-           bits = 0;
-           l = 0;
-           if (datumtype == 'B')
-               while (l++ < len) {
-                   if (utf8_source) {
-                       UV val = 0;
-                       NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
-                       bits |= val & 1;
-                   } else bits |= *str++ & 1;
-                   if (l & 7) bits <<= 1;
-                   else {
-                       PUSH_BYTE(utf8, cur, bits);
-                       bits = 0;
-                   }
-               }
-           else
-               /* datumtype == 'b' */
-               while (l++ < len) {
-                   if (utf8_source) {
-                       UV val = 0;
-                       NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
-                       if (val & 1) bits |= 0x80;
-                   } else if (*str++ & 1)
-                       bits |= 0x80;
-                   if (l & 7) bits >>= 1;
-                   else {
-                       PUSH_BYTE(utf8, cur, bits);
-                       bits = 0;
-                   }
-               }
-           l--;
-           if (l & 7) {
-               if (datumtype == 'B')
-                   bits <<= 7 - (l & 7);
-               else
-                   bits >>= 7 - (l & 7);
-               PUSH_BYTE(utf8, cur, bits);
-               l += 7;
-           }
-           /* Determine how many chars are left in the requested field */
-           l /= 8;
-           if (howlen == e_star) field_len = 0;
-           else field_len -= l;
-           Zero(cur, field_len, char);
-           cur += field_len;
-           break;
-       }
-       case 'H':
-       case 'h': {
-           const char *str, *end;
-           I32 l, field_len;
-           U8 bits;
-           bool utf8_source;
-           U32 utf8_flags;
-
-           fromstr = NEXTFROM;
-           str = SvPV_const(fromstr, fromlen);
-           end = str + fromlen;
-           if (DO_UTF8(fromstr)) {
-               utf8_source = TRUE;
-               utf8_flags  = warn_utf8 ? 0 : UTF8_ALLOW_ANY;
-           } else {
-               utf8_source = FALSE;
-               utf8_flags  = 0; /* Unused, but keep compilers happy */
-           }
-           if (howlen == e_star) len = fromlen;
-           field_len = (len+1)/2;
-           GROWING(utf8, cat, start, cur, field_len);
-           if (!utf8 && len > (I32)fromlen) len = fromlen;
-           bits = 0;
-           l = 0;
-           if (datumtype == 'H')
-               while (l++ < len) {
-                   if (utf8_source) {
-                       UV val = 0;
-                       NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
-                       if (val < 256 && isALPHA(val))
-                           bits |= (val + 9) & 0xf;
-                       else
-                           bits |= val & 0xf;
-                   } else if (isALPHA(*str))
-                       bits |= (*str++ + 9) & 0xf;
-                   else
-                       bits |= *str++ & 0xf;
-                   if (l & 1) bits <<= 4;
-                   else {
-                       PUSH_BYTE(utf8, cur, bits);
-                       bits = 0;
-                   }
-               }
-           else
-               while (l++ < len) {
-                   if (utf8_source) {
-                       UV val = 0;
-                       NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
-                       if (val < 256 && isALPHA(val))
-                           bits |= ((val + 9) & 0xf) << 4;
-                       else
-                           bits |= (val & 0xf) << 4;
-                   } else if (isALPHA(*str))
-                       bits |= ((*str++ + 9) & 0xf) << 4;
-                   else
-                       bits |= (*str++ & 0xf) << 4;
-                   if (l & 1) bits >>= 4;
-                   else {
-                       PUSH_BYTE(utf8, cur, bits);
-                       bits = 0;
-                   }
-               }
-           l--;
-           if (l & 1) {
-               PUSH_BYTE(utf8, cur, bits);
-               l++;
-           }
-           /* Determine how many chars are left in the requested field */
-           l /= 2;
-           if (howlen == e_star) field_len = 0;
-           else field_len -= l;
-           Zero(cur, field_len, char);
-           cur += field_len;
-           break;
-       }
-       case 'c':
-           while (len-- > 0) {
-               IV aiv;
-               fromstr = NEXTFROM;
-                if (SvNOK(fromstr) && Perl_isinfnan(SvNV(fromstr))) {
-                    /* 255 is a pretty arbitrary choice, but with
-                     * inf/-inf/nan and 256 bytes there is not much room. */
-                    aiv = 255;
-                   Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
-                                  "Character in 'c' format overflow in pack");
+                if (!utf8 && !SvUTF8(cat)) {
+                    marked_upgrade(aTHX_ cat, symptr);
+                    lookahead.flags |= FLAG_DO_UTF8;
+                    lookahead.strbeg = symptr->strbeg;
+                    utf8 = 1;
+                    start = SvPVX(cat);
+                    cur = start + SvCUR(cat);
                 }
-                else
-                    aiv = SvIV(fromstr);
-               if ((-128 > aiv || aiv > 127))
-                   Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
-                                  "Character in 'c' format wrapped in pack");
-               PUSH_BYTE(utf8, cur, (U8)(aiv & 0xff));
-           }
-           break;
-       case 'C':
-           if (len == 0) {
-               utf8 = (symptr->flags & FLAG_DO_UTF8) ? 1 : 0;
-               break;
-           }
-           while (len-- > 0) {
-               IV aiv;
-               fromstr = NEXTFROM;
-                if (SvNOK(fromstr) && Perl_isinfnan(SvNV(fromstr))) {
-                    /* See the 'c' case. */
-                    aiv = 255;
-                   Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
-                                  "Character in 'C' format overflow in pack");
+                if (howlen == e_star) {
+                    if (utf8) goto string_copy;
+                    len = fromlen+1;
+                }
+                s = aptr;
+                end = aptr + fromlen;
+                fromlen = datumtype == 'Z' ? len-1 : len;
+                while ((SSize_t) fromlen > 0 && s < end) {
+                    s += UTF8SKIP(s);
+                    fromlen--;
+                }
+                if (s > end)
+                    Perl_croak(aTHX_ "Malformed UTF-8 string in pack");
+                if (utf8) {
+                    len = fromlen;
+                    if (datumtype == 'Z') len++;
+                    fromlen = s-aptr;
+                    len += fromlen;
+
+                    goto string_copy;
+                }
+                fromlen = len - fromlen;
+                if (datumtype == 'Z') fromlen--;
+                if (howlen == e_star) {
+                    len = fromlen;
+                    if (datumtype == 'Z') len++;
+                }
+                GROWING(0, cat, start, cur, len);
+                if (!S_utf8_to_bytes(aTHX_ &aptr, end, cur, fromlen,
+                                  datumtype | TYPE_IS_PACK))
+                    Perl_croak(aTHX_ "panic: predicted utf8 length not available, "
+                               "for '%c', aptr=%p end=%p cur=%p, fromlen=%zu",
+                               (int)datumtype, aptr, end, cur, fromlen);
+                cur += fromlen;
+                len -= fromlen;
+            } else if (utf8) {
+                if (howlen == e_star) {
+                    len = fromlen;
+                    if (datumtype == 'Z') len++;
+                }
+                if (len <= (SSize_t) fromlen) {
+                    fromlen = len;
+                    if (datumtype == 'Z' && fromlen > 0) fromlen--;
                 }
+                /* assumes a byte expands to at most UTF8_EXPAND bytes on
+                   upgrade, so:
+                   expected_length <= from_len*UTF8_EXPAND + (len-from_len) */
+                GROWING(0, cat, start, cur, fromlen*(UTF8_EXPAND-1)+len);
+                len -= fromlen;
+                while (fromlen > 0) {
+                    cur = (char *) uvchr_to_utf8((U8 *) cur, * (U8 *) aptr);
+                    aptr++;
+                    fromlen--;
+                }
+            } else {
+              string_copy:
+                if (howlen == e_star) {
+                    len = fromlen;
+                    if (datumtype == 'Z') len++;
+                }
+                if (len <= (SSize_t) fromlen) {
+                    fromlen = len;
+                    if (datumtype == 'Z' && fromlen > 0) fromlen--;
+                }
+                GROWING(0, cat, start, cur, len);
+                Copy(aptr, cur, fromlen, char);
+                cur += fromlen;
+                len -= fromlen;
+            }
+            memset(cur, datumtype == 'A' ? ' ' : '\0', len);
+            cur += len;
+            SvTAINT(cat);
+            break;
+        }
+        case 'B':
+        case 'b': {
+            const char *str, *end;
+            SSize_t l, field_len;
+            U8 bits;
+            bool utf8_source;
+            U32 utf8_flags;
+
+            fromstr = NEXTFROM;
+            str = SvPV_const(fromstr, fromlen);
+            end = str + fromlen;
+            if (DO_UTF8(fromstr)) {
+                utf8_source = TRUE;
+                utf8_flags  = warn_utf8 ? 0 : UTF8_ALLOW_ANY;
+            } else {
+                utf8_source = FALSE;
+                utf8_flags  = 0; /* Unused, but keep compilers happy */
+            }
+            if (howlen == e_star) len = fromlen;
+            field_len = (len+7)/8;
+            GROWING(utf8, cat, start, cur, field_len);
+            if (len > (SSize_t)fromlen) len = fromlen;
+            bits = 0;
+            l = 0;
+            if (datumtype == 'B')
+                while (l++ < len) {
+                    if (utf8_source) {
+                        UV val = 0;
+                        NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
+                        bits |= val & 1;
+                    } else bits |= *str++ & 1;
+                    if (l & 7) bits <<= 1;
+                    else {
+                        PUSH_BYTE(utf8, cur, bits);
+                        bits = 0;
+                    }
+                }
+            else
+                /* datumtype == 'b' */
+                while (l++ < len) {
+                    if (utf8_source) {
+                        UV val = 0;
+                        NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
+                        if (val & 1) bits |= 0x80;
+                    } else if (*str++ & 1)
+                        bits |= 0x80;
+                    if (l & 7) bits >>= 1;
+                    else {
+                        PUSH_BYTE(utf8, cur, bits);
+                        bits = 0;
+                    }
+                }
+            l--;
+            if (l & 7) {
+                if (datumtype == 'B')
+                    bits <<= 7 - (l & 7);
                 else
-                    aiv = SvIV(fromstr);
-               if ((0 > aiv || aiv > 0xff))
-                   Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
-                                  "Character in 'C' format wrapped in pack");
-               PUSH_BYTE(utf8, cur, (U8)(aiv & 0xff));
-           }
-           break;
-       case 'W': {
-           char *end;
-           U8 in_bytes = (U8)IN_BYTES;
-
-           end = start+SvLEN(cat)-1;
-           if (utf8) end -= UTF8_MAXLEN-1;
-           while (len-- > 0) {
-               UV auv;
-               fromstr = NEXTFROM;
-               auv = SvUV(fromstr);
-               if (in_bytes) auv = auv % 0x100;
-               if (utf8) {
-                 W_utf8:
-                   if (cur > end) {
-                       *cur = '\0';
-                       SvCUR_set(cat, cur - start);
-
-                       GROWING(0, cat, start, cur, len+UTF8_MAXLEN);
-                       end = start+SvLEN(cat)-UTF8_MAXLEN;
-                   }
-                   cur = (char *) uvchr_to_utf8_flags((U8 *) cur,
-                                                      auv,
-                                                      warn_utf8 ?
-                                                      0 : UNICODE_ALLOW_ANY);
-               } else {
-                   if (auv >= 0x100) {
-                       if (!SvUTF8(cat)) {
-                           *cur = '\0';
-                           SvCUR_set(cat, cur - start);
-                           marked_upgrade(aTHX_ cat, symptr);
-                           lookahead.flags |= FLAG_DO_UTF8;
-                           lookahead.strbeg = symptr->strbeg;
-                           utf8 = 1;
-                           start = SvPVX(cat);
-                           cur = start + SvCUR(cat);
-                           end = start+SvLEN(cat)-UTF8_MAXLEN;
-                           goto W_utf8;
-                       }
-                       Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
-                                      "Character in 'W' format wrapped in pack");
-                       auv &= 0xff;
-                   }
-                   if (cur >= end) {
-                       *cur = '\0';
-                       SvCUR_set(cat, cur - start);
-                       GROWING(0, cat, start, cur, len+1);
-                       end = start+SvLEN(cat)-1;
-                   }
-                   *(U8 *) cur++ = (U8)auv;
-               }
-           }
-           break;
-       }
-       case 'U': {
-           char *end;
-
-           if (len == 0) {
-               if (!(symptr->flags & FLAG_DO_UTF8)) {
-                   marked_upgrade(aTHX_ cat, symptr);
-                   lookahead.flags |= FLAG_DO_UTF8;
-                   lookahead.strbeg = symptr->strbeg;
-               }
-               utf8 = 0;
-               goto no_change;
-           }
-
-           end = start+SvLEN(cat);
-           if (!utf8) end -= UTF8_MAXLEN;
-           while (len-- > 0) {
-               UV auv;
-               fromstr = NEXTFROM;
-               auv = SvUV(fromstr);
-               if (utf8) {
-                   U8 buffer[UTF8_MAXLEN], *endb;
-                   endb = uvchr_to_utf8_flags(buffer, auv,
-                                              warn_utf8 ?
-                                              0 : UNICODE_ALLOW_ANY);
-                   if (cur+(endb-buffer)*UTF8_EXPAND >= end) {
-                       *cur = '\0';
-                       SvCUR_set(cat, cur - start);
-                       GROWING(0, cat, start, cur,
-                               len+(endb-buffer)*UTF8_EXPAND);
-                       end = start+SvLEN(cat);
-                   }
-                    cur = S_bytes_to_uni(buffer, endb-buffer, cur, 0);
-               } else {
-                   if (cur >= end) {
-                       *cur = '\0';
-                       SvCUR_set(cat, cur - start);
-                       GROWING(0, cat, start, cur, len+UTF8_MAXLEN);
-                       end = start+SvLEN(cat)-UTF8_MAXLEN;
-                   }
-                   cur = (char *) uvchr_to_utf8_flags((U8 *) cur, auv,
-                                                      warn_utf8 ?
-                                                      0 : UNICODE_ALLOW_ANY);
-               }
-           }
-           break;
-       }
-       /* Float and double added by gnb@melba.bby.oz.au  22/11/89 */
-       case 'f':
-           while (len-- > 0) {
-               float afloat;
-               NV anv;
-               fromstr = NEXTFROM;
-               anv = SvNV(fromstr);
-# if defined(VMS) && !defined(_IEEE_FP)
-               /* IEEE fp overflow shenanigans are unavailable on VAX and optional
-                * on Alpha; fake it if we don't have them.
-                */
-               if (anv > FLT_MAX)
-                   afloat = FLT_MAX;
-               else if (anv < -FLT_MAX)
-                   afloat = -FLT_MAX;
-               else afloat = (float)anv;
+                    bits >>= 7 - (l & 7);
+                PUSH_BYTE(utf8, cur, bits);
+                l += 7;
+            }
+            /* Determine how many chars are left in the requested field */
+            l /= 8;
+            if (howlen == e_star) field_len = 0;
+            else field_len -= l;
+            Zero(cur, field_len, char);
+            cur += field_len;
+            break;
+        }
+        case 'H':
+        case 'h': {
+            const char *str, *end;
+            SSize_t l, field_len;
+            U8 bits;
+            bool utf8_source;
+            U32 utf8_flags;
+
+            fromstr = NEXTFROM;
+            str = SvPV_const(fromstr, fromlen);
+            end = str + fromlen;
+            if (DO_UTF8(fromstr)) {
+                utf8_source = TRUE;
+                utf8_flags  = warn_utf8 ? 0 : UTF8_ALLOW_ANY;
+            } else {
+                utf8_source = FALSE;
+                utf8_flags  = 0; /* Unused, but keep compilers happy */
+            }
+            if (howlen == e_star) len = fromlen;
+            field_len = (len+1)/2;
+            GROWING(utf8, cat, start, cur, field_len);
+            if (!utf8_source && len > (SSize_t)fromlen) len = fromlen;
+            bits = 0;
+            l = 0;
+            if (datumtype == 'H')
+                while (l++ < len) {
+                    if (utf8_source) {
+                        UV val = 0;
+                        NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
+                        if (val < 256 && isALPHA(val))
+                            bits |= (val + 9) & 0xf;
+                        else
+                            bits |= val & 0xf;
+                    } else if (isALPHA(*str))
+                        bits |= (*str++ + 9) & 0xf;
+                    else
+                        bits |= *str++ & 0xf;
+                    if (l & 1) bits <<= 4;
+                    else {
+                        PUSH_BYTE(utf8, cur, bits);
+                        bits = 0;
+                    }
+                }
+            else
+                while (l++ < len) {
+                    if (utf8_source) {
+                        UV val = 0;
+                        NEXT_UNI_VAL(val, cur, str, end, utf8_flags);
+                        if (val < 256 && isALPHA(val))
+                            bits |= ((val + 9) & 0xf) << 4;
+                        else
+                            bits |= (val & 0xf) << 4;
+                    } else if (isALPHA(*str))
+                        bits |= ((*str++ + 9) & 0xf) << 4;
+                    else
+                        bits |= (*str++ & 0xf) << 4;
+                    if (l & 1) bits >>= 4;
+                    else {
+                        PUSH_BYTE(utf8, cur, bits);
+                        bits = 0;
+                    }
+                }
+            l--;
+            if (l & 1) {
+                PUSH_BYTE(utf8, cur, bits);
+                l++;
+            }
+            /* Determine how many chars are left in the requested field */
+            l /= 2;
+            if (howlen == e_star) field_len = 0;
+            else field_len -= l;
+            Zero(cur, field_len, char);
+            cur += field_len;
+            break;
+        }
+        case 'c':
+            while (len-- > 0) {
+                IV aiv;
+                fromstr = NEXTFROM;
+                aiv = SvIV_no_inf(fromstr, datumtype);
+                if ((-128 > aiv || aiv > 127))
+                    Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
+                                   "Character in 'c' format wrapped in pack");
+                PUSH_BYTE(utf8, cur, (U8)(aiv & 0xff));
+            }
+            break;
+        case 'C':
+            if (len == 0) {
+                utf8 = (symptr->flags & FLAG_DO_UTF8) ? 1 : 0;
+                break;
+            }
+            while (len-- > 0) {
+                IV aiv;
+                fromstr = NEXTFROM;
+                aiv = SvIV_no_inf(fromstr, datumtype);
+                if ((0 > aiv || aiv > 0xff))
+                    Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
+                                   "Character in 'C' format wrapped in pack");
+                PUSH_BYTE(utf8, cur, (U8)(aiv & 0xff));
+            }
+            break;
+        case 'W': {
+            char *end;
+            U8 in_bytes = (U8)IN_BYTES;
+
+            end = start+SvLEN(cat)-1;
+            if (utf8) end -= UTF8_MAXLEN-1;
+            while (len-- > 0) {
+                UV auv;
+                fromstr = NEXTFROM;
+                auv = SvUV_no_inf(fromstr, datumtype);
+                if (in_bytes) auv = auv % 0x100;
+                if (utf8) {
+                  W_utf8:
+                    if (cur >= end) {
+                        *cur = '\0';
+                        SvCUR_set(cat, cur - start);
+
+                        GROWING(0, cat, start, cur, len+UTF8_MAXLEN);
+                        end = start+SvLEN(cat)-UTF8_MAXLEN;
+                    }
+                    cur = (char *) uvchr_to_utf8_flags((U8 *) cur, auv, 0);
+                } else {
+                    if (auv >= 0x100) {
+                        if (!SvUTF8(cat)) {
+                            *cur = '\0';
+                            SvCUR_set(cat, cur - start);
+                            marked_upgrade(aTHX_ cat, symptr);
+                            lookahead.flags |= FLAG_DO_UTF8;
+                            lookahead.strbeg = symptr->strbeg;
+                            utf8 = 1;
+                            start = SvPVX(cat);
+                            cur = start + SvCUR(cat);
+                            end = start+SvLEN(cat)-UTF8_MAXLEN;
+                            goto W_utf8;
+                        }
+                        Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
+                                       "Character in 'W' format wrapped in pack");
+                        auv &= 0xff;
+                    }
+                    if (cur >= end) {
+                        *cur = '\0';
+                        SvCUR_set(cat, cur - start);
+                        GROWING(0, cat, start, cur, len+1);
+                        end = start+SvLEN(cat)-1;
+                    }
+                    *(U8 *) cur++ = (U8)auv;
+                }
+            }
+            break;
+        }
+        case 'U': {
+            char *end;
+
+            if (len == 0) {
+                if (!(symptr->flags & FLAG_DO_UTF8)) {
+                    marked_upgrade(aTHX_ cat, symptr);
+                    lookahead.flags |= FLAG_DO_UTF8;
+                    lookahead.strbeg = symptr->strbeg;
+                }
+                utf8 = 0;
+                goto no_change;
+            }
+
+            end = start+SvLEN(cat);
+            if (!utf8) end -= UTF8_MAXLEN;
+            while (len-- > 0) {
+                UV auv;
+                fromstr = NEXTFROM;
+                auv = SvUV_no_inf(fromstr, datumtype);
+                if (utf8) {
+                    U8 buffer[UTF8_MAXLEN+1], *endb;
+                    endb = uvchr_to_utf8_flags(buffer, UNI_TO_NATIVE(auv), 0);
+                    if (cur+(endb-buffer)*UTF8_EXPAND >= end) {
+                        *cur = '\0';
+                        SvCUR_set(cat, cur - start);
+                        GROWING(0, cat, start, cur,
+                                len+(endb-buffer)*UTF8_EXPAND);
+                        end = start+SvLEN(cat);
+                    }
+                    cur = my_bytes_to_utf8(buffer, endb-buffer, cur, 0);
+                } else {
+                    if (cur >= end) {
+                        *cur = '\0';
+                        SvCUR_set(cat, cur - start);
+                        GROWING(0, cat, start, cur, len+UTF8_MAXLEN);
+                        end = start+SvLEN(cat)-UTF8_MAXLEN;
+                    }
+                    cur = (char *) uvchr_to_utf8_flags((U8 *) cur,
+                                                       UNI_TO_NATIVE(auv),
+                                                       0);
+                }
+            }
+            break;
+        }
+        /* Float and double added by gnb@melba.bby.oz.au  22/11/89 */
+        case 'f':
+            while (len-- > 0) {
+                float afloat;
+                NV anv;
+                fromstr = NEXTFROM;
+                anv = SvNV(fromstr);
+# if (defined(VMS) && !defined(_IEEE_FP)) || defined(DOUBLE_IS_VAX_FLOAT)
+                /* IEEE fp overflow shenanigans are unavailable on VAX and optional
+                 * on Alpha; fake it if we don't have them.
+                 */
+                if (anv > FLT_MAX)
+                    afloat = FLT_MAX;
+                else if (anv < -FLT_MAX)
+                    afloat = -FLT_MAX;
+                else afloat = (float)anv;
 # else
-               afloat = (float)anv;
+#  if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
+                if(Perl_isnan(anv))
+                    afloat = (float)NV_NAN;
+                else
+#  endif
+#  ifdef NV_INF
+                /* a simple cast to float is undefined if outside
+                 * the range of values that can be represented */
+                afloat = (float)(anv >  FLT_MAX ?  NV_INF :
+                                 anv < -FLT_MAX ? -NV_INF : anv);
+#  endif
 # endif
                 PUSH_VAR(utf8, cur, afloat, needs_swap);
-           }
-           break;
-       case 'd':
-           while (len-- > 0) {
-               double adouble;
-               NV anv;
-               fromstr = NEXTFROM;
-               anv = SvNV(fromstr);
-# if defined(VMS) && !defined(_IEEE_FP)
-               /* IEEE fp overflow shenanigans are unavailable on VAX and optional
-                * on Alpha; fake it if we don't have them.
-                */
-               if (anv > DBL_MAX)
-                   adouble = DBL_MAX;
-               else if (anv < -DBL_MAX)
-                   adouble = -DBL_MAX;
-               else adouble = (double)anv;
+            }
+            break;
+        case 'd':
+            while (len-- > 0) {
+                double adouble;
+                NV anv;
+                fromstr = NEXTFROM;
+                anv = SvNV(fromstr);
+# if (defined(VMS) && !defined(_IEEE_FP)) || defined(DOUBLE_IS_VAX_FLOAT)
+                /* IEEE fp overflow shenanigans are unavailable on VAX and optional
+                 * on Alpha; fake it if we don't have them.
+                 */
+                if (anv > DBL_MAX)
+                    adouble = DBL_MAX;
+                else if (anv < -DBL_MAX)
+                    adouble = -DBL_MAX;
+                else adouble = (double)anv;
 # else
-               adouble = (double)anv;
+                adouble = (double)anv;
 # endif
                 PUSH_VAR(utf8, cur, adouble, needs_swap);
-           }
-           break;
-       case 'F': {
-           NV_bytes anv;
-           Zero(&anv, 1, NV); /* can be long double with unused bits */
-           while (len-- > 0) {
-               fromstr = NEXTFROM;
+            }
+            break;
+        case 'F': {
+            NV_bytes anv;
+            Zero(&anv, 1, NV); /* can be long double with unused bits */
+            while (len-- > 0) {
+                fromstr = NEXTFROM;
 #ifdef __GNUC__
-               /* to work round a gcc/x86 bug; don't use SvNV */
-               anv.nv = sv_2nv(fromstr);
+                /* to work round a gcc/x86 bug; don't use SvNV */
+                anv.nv = sv_2nv(fromstr);
+#    if defined(LONGDOUBLE_X86_80_BIT) && defined(USE_LONG_DOUBLE) \
+         && LONG_DOUBLESIZE > 10
+                /* GCC sometimes overwrites the padding in the
+                   assignment above */
+                Zero(anv.bytes+10, sizeof(anv.bytes) - 10, U8);
+#    endif
 #else
-               anv.nv = SvNV(fromstr);
+                anv.nv = SvNV(fromstr);
 #endif
                 PUSH_BYTES(utf8, cur, anv.bytes, sizeof(anv.bytes), needs_swap);
-           }
-           break;
-       }
-#if defined(HAS_LONG_DOUBLE) && defined(USE_LONG_DOUBLE)
-       case 'D': {
-           ld_bytes aldouble;
-           /* long doubles can have unused bits, which may be nonzero */
-           Zero(&aldouble, 1, long double);
-           while (len-- > 0) {
-               fromstr = NEXTFROM;
+            }
+            break;
+        }
+#if defined(HAS_LONG_DOUBLE)
+        case 'D': {
+            ld_bytes aldouble;
+            /* long doubles can have unused bits, which may be nonzero */
+            Zero(&aldouble, 1, long double);
+            while (len-- > 0) {
+                fromstr = NEXTFROM;
 #  ifdef __GNUC__
-               /* to work round a gcc/x86 bug; don't use SvNV */
-               aldouble.ld = (long double)sv_2nv(fromstr);
+                /* to work round a gcc/x86 bug; don't use SvNV */
+                aldouble.ld = (long double)sv_2nv(fromstr);
+#    if defined(LONGDOUBLE_X86_80_BIT) && LONG_DOUBLESIZE > 10
+                /* GCC sometimes overwrites the padding in the
+                   assignment above */
+                Zero(aldouble.bytes+10, sizeof(aldouble.bytes) - 10, U8);
+#    endif
 #  else
-               aldouble.ld = (long double)SvNV(fromstr);
+                aldouble.ld = (long double)SvNV(fromstr);
 #  endif
                 PUSH_BYTES(utf8, cur, aldouble.bytes, sizeof(aldouble.bytes),
                            needs_swap);
-           }
-           break;
-       }
+            }
+            break;
+        }
 #endif
-       case 'n' | TYPE_IS_SHRIEKING:
-       case 'n':
-           while (len-- > 0) {
-               I16 ai16;
-               fromstr = NEXTFROM;
-               ai16 = (I16)SvIV(fromstr);
-               ai16 = PerlSock_htons(ai16);
+        case 'n' | TYPE_IS_SHRIEKING:
+        case 'n':
+            while (len-- > 0) {
+                I16 ai16;
+                fromstr = NEXTFROM;
+                ai16 = (I16)SvIV_no_inf(fromstr, datumtype);
+                ai16 = PerlSock_htons(ai16);
                 PUSH16(utf8, cur, &ai16, FALSE);
-           }
-           break;
-       case 'v' | TYPE_IS_SHRIEKING:
-       case 'v':
-           while (len-- > 0) {
-               I16 ai16;
-               fromstr = NEXTFROM;
-               ai16 = (I16)SvIV(fromstr);
-               ai16 = htovs(ai16);
+            }
+            break;
+        case 'v' | TYPE_IS_SHRIEKING:
+        case 'v':
+            while (len-- > 0) {
+                I16 ai16;
+                fromstr = NEXTFROM;
+                ai16 = (I16)SvIV_no_inf(fromstr, datumtype);
+                ai16 = htovs(ai16);
                 PUSH16(utf8, cur, &ai16, FALSE);
-           }
-           break;
+            }
+            break;
         case 'S' | TYPE_IS_SHRIEKING:
 #if SHORTSIZE != SIZE16
-           while (len-- > 0) {
-               unsigned short aushort;
-               fromstr = NEXTFROM;
-               aushort = SvUV(fromstr);
+            while (len-- > 0) {
+                unsigned short aushort;
+                fromstr = NEXTFROM;
+                aushort = SvUV_no_inf(fromstr, datumtype);
                 PUSH_VAR(utf8, cur, aushort, needs_swap);
-           }
+            }
             break;
 #else
             /* FALLTHROUGH */
 #endif
-       case 'S':
-           while (len-- > 0) {
-               U16 au16;
-               fromstr = NEXTFROM;
-               au16 = (U16)SvUV(fromstr);
+        case 'S':
+            while (len-- > 0) {
+                U16 au16;
+                fromstr = NEXTFROM;
+                au16 = (U16)SvUV_no_inf(fromstr, datumtype);
                 PUSH16(utf8, cur, &au16, needs_swap);
-           }
-           break;
-       case 's' | TYPE_IS_SHRIEKING:
+            }
+            break;
+        case 's' | TYPE_IS_SHRIEKING:
 #if SHORTSIZE != SIZE16
-           while (len-- > 0) {
-               short ashort;
-               fromstr = NEXTFROM;
-               ashort = SvIV(fromstr);
+            while (len-- > 0) {
+                short ashort;
+                fromstr = NEXTFROM;
+                ashort = SvIV_no_inf(fromstr, datumtype);
                 PUSH_VAR(utf8, cur, ashort, needs_swap);
-           }
+            }
             break;
 #else
             /* FALLTHROUGH */
 #endif
-       case 's':
-           while (len-- > 0) {
-               I16 ai16;
-               fromstr = NEXTFROM;
-               ai16 = (I16)SvIV(fromstr);
+        case 's':
+            while (len-- > 0) {
+                I16 ai16;
+                fromstr = NEXTFROM;
+                ai16 = (I16)SvIV_no_inf(fromstr, datumtype);
                 PUSH16(utf8, cur, &ai16, needs_swap);
-           }
-           break;
-       case 'I':
-       case 'I' | TYPE_IS_SHRIEKING:
-           while (len-- > 0) {
-               unsigned int auint;
-               fromstr = NEXTFROM;
-               auint = SvUV(fromstr);
+            }
+            break;
+        case 'I':
+        case 'I' | TYPE_IS_SHRIEKING:
+            while (len-- > 0) {
+                unsigned int auint;
+                fromstr = NEXTFROM;
+                auint = SvUV_no_inf(fromstr, datumtype);
                 PUSH_VAR(utf8, cur, auint, needs_swap);
-           }
-           break;
-       case 'j':
-           while (len-- > 0) {
-               IV aiv;
-               fromstr = NEXTFROM;
-               aiv = SvIV(fromstr);
+            }
+            break;
+        case 'j':
+            while (len-- > 0) {
+                IV aiv;
+                fromstr = NEXTFROM;
+                aiv = SvIV_no_inf(fromstr, datumtype);
                 PUSH_VAR(utf8, cur, aiv, needs_swap);
-           }
-           break;
-       case 'J':
-           while (len-- > 0) {
-               UV auv;
-               fromstr = NEXTFROM;
-               auv = SvUV(fromstr);
+            }
+            break;
+        case 'J':
+            while (len-- > 0) {
+                UV auv;
+                fromstr = NEXTFROM;
+                auv = SvUV_no_inf(fromstr, datumtype);
                 PUSH_VAR(utf8, cur, auv, needs_swap);
-           }
-           break;
-       case 'w':
+            }
+            break;
+        case 'w':
             while (len-- > 0) {
-               NV anv;
-               fromstr = NEXTFROM;
-               anv = SvNV(fromstr);
-
-               if (anv < 0) {
-                   *cur = '\0';
-                   SvCUR_set(cat, cur - start);
-                   Perl_croak(aTHX_ "Cannot compress negative numbers in pack");
-               }
+                NV anv;
+                fromstr = NEXTFROM;
+                S_sv_check_infnan(aTHX_ fromstr, datumtype);
+                anv = SvNV_nomg(fromstr);
+
+                if (anv < 0) {
+                    *cur = '\0';
+                    SvCUR_set(cat, cur - start);
+                    Perl_croak(aTHX_ "Cannot compress negative numbers in pack");
+                }
 
                 /* 0xFFFFFFFFFFFFFFFF may cast to 18446744073709551616.0,
                    which is == UV_MAX_P1. IOK is fine (instead of UV_only), as
                    any negative IVs will have already been got by the croak()
                    above. IOK is untrue for fractions, so we test them
                    against UV_MAX_P1.  */
-               if (SvIOK(fromstr) || anv < UV_MAX_P1) {
-                   char   buf[(sizeof(UV)*CHAR_BIT)/7+1];
-                   char  *in = buf + sizeof(buf);
-                   UV     auv = SvUV(fromstr);
-
-                   do {
-                       *--in = (char)((auv & 0x7f) | 0x80);
-                       auv >>= 7;
-                   } while (auv);
-                   buf[sizeof(buf) - 1] &= 0x7f; /* clear continue bit */
-                   PUSH_GROWING_BYTES(utf8, cat, start, cur,
-                                      in, (buf + sizeof(buf)) - in);
-               } else if (SvPOKp(fromstr))
-                   goto w_string;
-               else if (SvNOKp(fromstr)) {
-                   /* 10**NV_MAX_10_EXP is the largest power of 10
-                      so 10**(NV_MAX_10_EXP+1) is definitely unrepresentable
-                      given 10**(NV_MAX_10_EXP+1) == 128 ** x solve for x:
-                      x = (NV_MAX_10_EXP+1) * log (10) / log (128)
-                      And with that many bytes only Inf can overflow.
-                      Some C compilers are strict about integral constant
-                      expressions so we conservatively divide by a slightly
-                      smaller integer instead of multiplying by the exact
-                      floating-point value.
-                   */
+                if (SvIOK(fromstr) || anv < UV_MAX_P1) {
+                    char   buf[(sizeof(UV)*CHAR_BIT)/7+1];
+                    char  *in = buf + sizeof(buf);
+                    UV     auv = SvUV_nomg(fromstr);
+
+                    do {
+                        *--in = (char)((auv & 0x7f) | 0x80);
+                        auv >>= 7;
+                    } while (auv);
+                    buf[sizeof(buf) - 1] &= 0x7f; /* clear continue bit */
+                    PUSH_GROWING_BYTES(utf8, cat, start, cur,
+                                       in, (buf + sizeof(buf)) - in);
+                } else if (SvPOKp(fromstr))
+                    goto w_string;
+                else if (SvNOKp(fromstr)) {
+                    /* 10**NV_MAX_10_EXP is the largest power of 10
+                       so 10**(NV_MAX_10_EXP+1) is definitely unrepresentable
+                       given 10**(NV_MAX_10_EXP+1) == 128 ** x solve for x:
+                       x = (NV_MAX_10_EXP+1) * log (10) / log (128)
+                       And with that many bytes only Inf can overflow.
+                       Some C compilers are strict about integral constant
+                       expressions so we conservatively divide by a slightly
+                       smaller integer instead of multiplying by the exact
+                       floating-point value.
+                    */
 #ifdef NV_MAX_10_EXP
-                   /* char   buf[1 + (int)((NV_MAX_10_EXP + 1) * 0.47456)]; -- invalid C */
-                   char   buf[1 + (int)((NV_MAX_10_EXP + 1) / 2)]; /* valid C */
+                    /* char   buf[1 + (int)((NV_MAX_10_EXP + 1) * 0.47456)]; -- invalid C */
+                    char   buf[1 + (int)((NV_MAX_10_EXP + 1) / 2)]; /* valid C */
 #else
-                   /* char   buf[1 + (int)((308 + 1) * 0.47456)]; -- invalid C */
-                   char   buf[1 + (int)((308 + 1) / 2)]; /* valid C */
+                    /* char   buf[1 + (int)((308 + 1) * 0.47456)]; -- invalid C */
+                    char   buf[1 + (int)((308 + 1) / 2)]; /* valid C */
 #endif
-                   char  *in = buf + sizeof(buf);
-                    static const char S_cannot_compress[] =
-                        "Cannot compress integer in pack";
-
-                    if (Perl_isinfnan(anv))
-                        Perl_croak(aTHX_ S_cannot_compress);
-
-                   anv = Perl_floor(anv);
-                   do {
-                       const NV next = Perl_floor(anv / 128);
-                       if (in <= buf)  /* this cannot happen ;-) */
-                           Perl_croak(aTHX_ S_cannot_compress);
-                       *--in = (unsigned char)(anv - (next * 128)) | 0x80;
-                       anv = next;
-                   } while (anv > 0);
-                   buf[sizeof(buf) - 1] &= 0x7f; /* clear continue bit */
-                   PUSH_GROWING_BYTES(utf8, cat, start, cur,
-                                      in, (buf + sizeof(buf)) - in);
-               } else {
-                   const char     *from;
-                   char           *result, *in;
-                   SV             *norm;
-                   STRLEN          len;
-                   bool            done;
-
-                 w_string:
-                   /* Copy string and check for compliance */
-                   from = SvPV_const(fromstr, len);
-                   if ((norm = is_an_int(from, len)) == NULL)
-                       Perl_croak(aTHX_ "Can only compress unsigned integers in pack");
-
-                   Newx(result, len, char);
-                   in = result + len;
-                   done = FALSE;
-                   while (!done) *--in = div128(norm, &done) | 0x80;
-                   result[len - 1] &= 0x7F; /* clear continue bit */
-                   PUSH_GROWING_BYTES(utf8, cat, start, cur,
-                                      in, (result + len) - in);
-                   Safefree(result);
-                   SvREFCNT_dec(norm); /* free norm */
-               }
-           }
-            break;
-       case 'i':
-       case 'i' | TYPE_IS_SHRIEKING:
-           while (len-- > 0) {
-               int aint;
-               fromstr = NEXTFROM;
-               aint = SvIV(fromstr);
+                    char  *in = buf + sizeof(buf);
+
+                    anv = Perl_floor(anv);
+                    do {
+                        const NV next = Perl_floor(anv / 128);
+                        if (in <= buf)  /* this cannot happen ;-) */
+                            Perl_croak(aTHX_ "Cannot compress integer in pack");
+                        *--in = (unsigned char)(anv - (next * 128)) | 0x80;
+                        anv = next;
+                    } while (anv > 0);
+                    buf[sizeof(buf) - 1] &= 0x7f; /* clear continue bit */
+                    PUSH_GROWING_BYTES(utf8, cat, start, cur,
+                                       in, (buf + sizeof(buf)) - in);
+                } else {
+                    const char     *from;
+                    char           *result, *in;
+                    SV             *norm;
+                    STRLEN          len;
+                    bool            done;
+
+                  w_string:
+                    /* Copy string and check for compliance */
+                    from = SvPV_nomg_const(fromstr, len);
+                    if ((norm = is_an_int(from, len)) == NULL)
+                        Perl_croak(aTHX_ "Can only compress unsigned integers in pack");
+
+                    Newx(result, len, char);
+                    in = result + len;
+                    done = FALSE;
+                    while (!done) *--in = div128(norm, &done) | 0x80;
+                    result[len - 1] &= 0x7F; /* clear continue bit */
+                    PUSH_GROWING_BYTES(utf8, cat, start, cur,
+                                       in, (result + len) - in);
+                    Safefree(result);
+                    SvREFCNT_dec(norm);        /* free norm */
+                }
+            }
+            break;
+        case 'i':
+        case 'i' | TYPE_IS_SHRIEKING:
+            while (len-- > 0) {
+                int aint;
+                fromstr = NEXTFROM;
+                aint = SvIV_no_inf(fromstr, datumtype);
                 PUSH_VAR(utf8, cur, aint, needs_swap);
-           }
-           break;
-       case 'N' | TYPE_IS_SHRIEKING:
-       case 'N':
-           while (len-- > 0) {
-               U32 au32;
-               fromstr = NEXTFROM;
-               au32 = SvUV(fromstr);
-               au32 = PerlSock_htonl(au32);
+            }
+            break;
+        case 'N' | TYPE_IS_SHRIEKING:
+        case 'N':
+            while (len-- > 0) {
+                U32 au32;
+                fromstr = NEXTFROM;
+                au32 = SvUV_no_inf(fromstr, datumtype);
+                au32 = PerlSock_htonl(au32);
                 PUSH32(utf8, cur, &au32, FALSE);
-           }
-           break;
-       case 'V' | TYPE_IS_SHRIEKING:
-       case 'V':
-           while (len-- > 0) {
-               U32 au32;
-               fromstr = NEXTFROM;
-               au32 = SvUV(fromstr);
-               au32 = htovl(au32);
+            }
+            break;
+        case 'V' | TYPE_IS_SHRIEKING:
+        case 'V':
+            while (len-- > 0) {
+                U32 au32;
+                fromstr = NEXTFROM;
+                au32 = SvUV_no_inf(fromstr, datumtype);
+                au32 = htovl(au32);
                 PUSH32(utf8, cur, &au32, FALSE);
-           }
-           break;
-       case 'L' | TYPE_IS_SHRIEKING:
+            }
+            break;
+        case 'L' | TYPE_IS_SHRIEKING:
 #if LONGSIZE != SIZE32
-           while (len-- > 0) {
-               unsigned long aulong;
-               fromstr = NEXTFROM;
-               aulong = SvUV(fromstr);
+            while (len-- > 0) {
+                unsigned long aulong;
+                fromstr = NEXTFROM;
+                aulong = SvUV_no_inf(fromstr, datumtype);
                 PUSH_VAR(utf8, cur, aulong, needs_swap);
-           }
-           break;
+            }
+            break;
 #else
             /* Fall though! */
 #endif
-       case 'L':
-           while (len-- > 0) {
-               U32 au32;
-               fromstr = NEXTFROM;
-               au32 = SvUV(fromstr);
+        case 'L':
+            while (len-- > 0) {
+                U32 au32;
+                fromstr = NEXTFROM;
+                au32 = SvUV_no_inf(fromstr, datumtype);
                 PUSH32(utf8, cur, &au32, needs_swap);
-           }
-           break;
-       case 'l' | TYPE_IS_SHRIEKING:
+            }
+            break;
+        case 'l' | TYPE_IS_SHRIEKING:
 #if LONGSIZE != SIZE32
-           while (len-- > 0) {
-               long along;
-               fromstr = NEXTFROM;
-               along = SvIV(fromstr);
+            while (len-- > 0) {
+                long along;
+                fromstr = NEXTFROM;
+                along = SvIV_no_inf(fromstr, datumtype);
                 PUSH_VAR(utf8, cur, along, needs_swap);
-           }
-           break;
+            }
+            break;
 #else
             /* Fall though! */
 #endif
-       case 'l':
+        case 'l':
             while (len-- > 0) {
-               I32 ai32;
-               fromstr = NEXTFROM;
-               ai32 = SvIV(fromstr);
+                I32 ai32;
+                fromstr = NEXTFROM;
+                ai32 = SvIV_no_inf(fromstr, datumtype);
                 PUSH32(utf8, cur, &ai32, needs_swap);
-           }
-           break;
+            }
+            break;
 #if defined(HAS_QUAD) && IVSIZE >= 8
-       case 'Q':
-           while (len-- > 0) {
-               Uquad_t auquad;
-               fromstr = NEXTFROM;
-               auquad = (Uquad_t) SvUV(fromstr);
+        case 'Q':
+            while (len-- > 0) {
+                Uquad_t auquad;
+                fromstr = NEXTFROM;
+                auquad = (Uquad_t) SvUV_no_inf(fromstr, datumtype);
                 PUSH_VAR(utf8, cur, auquad, needs_swap);
-           }
-           break;
-       case 'q':
-           while (len-- > 0) {
-               Quad_t aquad;
-               fromstr = NEXTFROM;
-               aquad = (Quad_t)SvIV(fromstr);
+            }
+            break;
+        case 'q':
+            while (len-- > 0) {
+                Quad_t aquad;
+                fromstr = NEXTFROM;
+                aquad = (Quad_t)SvIV_no_inf(fromstr, datumtype);
                 PUSH_VAR(utf8, cur, aquad, needs_swap);
-           }
-           break;
+            }
+            break;
 #endif
-       case 'P':
-           len = 1;            /* assume SV is correct length */
-           GROWING(utf8, cat, start, cur, sizeof(char *));
-           /* FALLTHROUGH */
-       case 'p':
-           while (len-- > 0) {
-               const char *aptr;
-
-               fromstr = NEXTFROM;
-               SvGETMAGIC(fromstr);
-               if (!SvOK(fromstr)) aptr = NULL;
-               else {
-                   /* XXX better yet, could spirit away the string to
-                    * a safe spot and hang on to it until the result
-                    * of pack() (and all copies of the result) are
-                    * gone.
-                    */
-                   if ((SvTEMP(fromstr) || (SvPADTMP(fromstr) &&
-                            !SvREADONLY(fromstr)))) {
-                       Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
-                                      "Attempt to pack pointer to temporary value");
-                   }
-                   if (SvPOK(fromstr) || SvNIOK(fromstr))
-                       aptr = SvPV_nomg_const_nolen(fromstr);
-                   else
-                       aptr = SvPV_force_flags_nolen(fromstr, 0);
-               }
+        case 'P':
+            len = 1;           /* assume SV is correct length */
+            GROWING(utf8, cat, start, cur, sizeof(char *));
+            /* FALLTHROUGH */
+        case 'p':
+            while (len-- > 0) {
+                const char *aptr;
+
+                fromstr = NEXTFROM;
+                SvGETMAGIC(fromstr);
+                if (!SvOK(fromstr)) aptr = NULL;
+                else {
+                    /* XXX better yet, could spirit away the string to
+                     * a safe spot and hang on to it until the result
+                     * of pack() (and all copies of the result) are
+                     * gone.
+                     */
+                    if (((SvTEMP(fromstr) && SvREFCNT(fromstr) == 1)
+                         || (SvPADTMP(fromstr) &&
+                             !SvREADONLY(fromstr)))) {
+                        Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
+                                       "Attempt to pack pointer to temporary value");
+                    }
+                    if (SvPOK(fromstr) || SvNIOK(fromstr))
+                        aptr = SvPV_nomg_const_nolen(fromstr);
+                    else
+                        aptr = SvPV_force_flags_nolen(fromstr, 0);
+                }
                 PUSH_VAR(utf8, cur, aptr, needs_swap);
-           }
-           break;
-       case 'u': {
-           const char *aptr, *aend;
-           bool from_utf8;
-
-           fromstr = NEXTFROM;
-           if (len <= 2) len = 45;
-           else len = len / 3 * 3;
-           if (len >= 64) {
-               Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
-                              "Field too wide in 'u' format in pack");
-               len = 63;
-           }
-           aptr = SvPV_const(fromstr, fromlen);
-           from_utf8 = DO_UTF8(fromstr);
-           if (from_utf8) {
-               aend = aptr + fromlen;
-               fromlen = sv_len_utf8_nomg(fromstr);
-           } else aend = NULL; /* Unused, but keep compilers happy */
-           GROWING(utf8, cat, start, cur, (fromlen+2) / 3 * 4 + (fromlen+len-1)/len * 2);
-           while (fromlen > 0) {
-               U8 *end;
-               I32 todo;
-               U8 hunk[1+63/3*4+1];
-
-               if ((I32)fromlen > len)
-                   todo = len;
-               else
-                   todo = fromlen;
-               if (from_utf8) {
-                   char buffer[64];
-                   if (!uni_to_bytes(aTHX_ &aptr, aend, buffer, todo,
-                                     'u' | TYPE_IS_PACK)) {
-                       *cur = '\0';
-                       SvCUR_set(cat, cur - start);
-                       Perl_croak(aTHX_ "panic: string is shorter than advertised, "
-                                  "aptr=%p, aend=%p, buffer=%p, todo=%ld",
-                                  aptr, aend, buffer, (long) todo);
-                   }
-                   end = doencodes(hunk, buffer, todo);
-               } else {
-                   end = doencodes(hunk, aptr, todo);
-                   aptr += todo;
-               }
-               PUSH_BYTES(utf8, cur, hunk, end-hunk, 0);
-               fromlen -= todo;
-           }
-           break;
-       }
-       }
-       *cur = '\0';
-       SvCUR_set(cat, cur - start);
+            }
+            break;
+        case 'u': {
+            const char *aptr, *aend;
+            bool from_utf8;
+
+            fromstr = NEXTFROM;
+            if (len <= 2) len = 45;
+            else len = len / 3 * 3;
+            if (len >= 64) {
+                Perl_ck_warner(aTHX_ packWARN(WARN_PACK),
+                               "Field too wide in 'u' format in pack");
+                len = 63;
+            }
+            aptr = SvPV_const(fromstr, fromlen);
+            from_utf8 = DO_UTF8(fromstr);
+            if (from_utf8) {
+                aend = aptr + fromlen;
+                fromlen = sv_len_utf8_nomg(fromstr);
+            } else aend = NULL; /* Unused, but keep compilers happy */
+            GROWING(utf8, cat, start, cur, (fromlen+2) / 3 * 4 + (fromlen+len-1)/len * 2);
+            while (fromlen > 0) {
+                U8 *end;
+                SSize_t todo;
+                U8 hunk[1+63/3*4+1];
+
+                if ((SSize_t)fromlen > len)
+                    todo = len;
+                else
+                    todo = fromlen;
+                if (from_utf8) {
+                    char buffer[64];
+                    if (!S_utf8_to_bytes(aTHX_ &aptr, aend, buffer, todo,
+                                      'u' | TYPE_IS_PACK)) {
+                        *cur = '\0';
+                        SvCUR_set(cat, cur - start);
+                        Perl_croak(aTHX_ "panic: string is shorter than advertised, "
+                                   "aptr=%p, aend=%p, buffer=%p, todo=%zd",
+                                   aptr, aend, buffer, todo);
+                    }
+                    end = doencodes(hunk, (const U8 *)buffer, todo);
+                } else {
+                    end = doencodes(hunk, (const U8 *)aptr, todo);
+                    aptr += todo;
+                }
+                PUSH_BYTES(utf8, cur, hunk, end-hunk, 0);
+                fromlen -= todo;
+            }
+            break;
+        }
+        }
+        *cur = '\0';
+        SvCUR_set(cat, cur - start);
       no_change:
-       *symptr = lookahead;
+        *symptr = lookahead;
     }
     return beglist;
 }
@@ -3128,11 +3143,26 @@ PP(pp_pack)
     const char *patend = pat + fromlen;
 
     MARK++;
-    sv_setpvs(cat, "");
+    SvPVCLEAR(cat);
     SvUTF8_off(cat);
 
     packlist(cat, pat, patend, MARK, SP + 1);
 
+    if (SvUTF8(cat)) {
+        STRLEN result_len;
+        const char * result = SvPV_nomg(cat, result_len);
+        const U8 * error_pos;
+
+        if (! is_utf8_string_loc((U8 *) result, result_len, &error_pos)) {
+            _force_out_malformed_utf8_message(error_pos,
+                                              (U8 *) result + result_len,
+                                              0, /* no flags */
+                                              1 /* Die */
+                                            );
+            NOT_REACHED; /* NOTREACHED */
+        }
+    }
+
     SvSETMAGIC(cat);
     SP = ORIGMARK;
     PUSHs(cat);
@@ -3140,11 +3170,5 @@ PP(pp_pack)
 }
 
 /*
- * Local variables:
- * c-indentation-style: bsd
- * c-basic-offset: 4
- * indent-tabs-mode: nil
- * End:
- *
  * ex: set ts=8 sts=4 sw=4 et:
  */