Add regnode EXACTFU_ONLY8
authorKarl Williamson <khw@cpan.org>
Tue, 27 Nov 2018 16:42:45 +0000 (09:42 -0700)
committerKarl Williamson <khw@cpan.org>
Tue, 27 Nov 2018 16:54:46 +0000 (09:54 -0700)
This is a regnode that otherwise would be an EXACTFU except that it
contains a code point that requires UTF-8 to match, including all the
possible folds involving it.  Hence if the target string isn't UTF-8, we
know it can't possibly match, without needing to try.

For completeness, there could also be an EXACTFAA_ONLY8 and an
EXACTFL_ONLY8 created, but I think these are unlikely to actually appear
in the wild, since using /aa is mainly about ASCII, and /l mostly will
involve characters that don't require UTF-8.

pod/perldebguts.pod
regcomp.c
regcomp.sym
regexec.c
regnodes.h

index e625d02..191c875 100644 (file)
@@ -676,6 +676,8 @@ will be lost.
 
  EXACT_ONLY8      str        Like EXACT, but no strings that aren't in
                              UTF-8 can match
+ EXACTFU_ONLY8    str        Like EXACTFU, but no strings that aren't in
+                             UTF-8 can match
 
  # Do nothing types
 
index 6179c69..7b1b494 100644 (file)
--- a/regcomp.c
+++ b/regcomp.c
@@ -2741,7 +2741,8 @@ S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch,
         if (    noper < tail
             && (    OP(noper) == flags
                 || (flags == EXACT && OP(noper) == EXACT_ONLY8)
-                || (flags == EXACTFU && OP(noper) == EXACTFU_SS)))
+                || (flags == EXACTFU && (   OP(noper) == EXACTFU_ONLY8
+                                         || OP(noper) == EXACTFU_SS))) )
         {
             uc= (U8*)STRING(noper);
             e= uc + STR_LEN(noper);
@@ -2958,7 +2959,8 @@ S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch,
             if (    noper < tail
                 && (    OP(noper) == flags
                     || (flags == EXACT && OP(noper) == EXACT_ONLY8)
-                    || (flags == EXACTFU && OP(noper) == EXACTFU_SS) ) )
+                    || (flags == EXACTFU && (   OP(noper) == EXACTFU_ONLY8
+                                             || OP(noper) == EXACTFU_SS))) )
             {
                 const U8 *uc= (U8*)STRING(noper);
                 const U8 *e= uc + STR_LEN(noper);
@@ -3182,7 +3184,8 @@ S_make_trie(pTHX_ RExC_state_t *pRExC_state, regnode *startbranch,
             if (    noper < tail
                 && (    OP(noper) == flags
                     || (flags == EXACT && OP(noper) == EXACT_ONLY8)
-                    || (flags == EXACTFU && OP(noper) == EXACTFU_SS) ) )
+                    || (flags == EXACTFU && (   OP(noper) == EXACTFU_ONLY8
+                                             || OP(noper) == EXACTFU_SS))) )
             {
                 const U8 *uc= (U8*)STRING(noper);
                 const U8 *e= uc + STR_LEN(noper);
@@ -4671,6 +4674,7 @@ S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
                                 EXACT           | EXACT
                                 EXACT_ONLY8     | EXACT
                                 EXACTFU         | EXACTFU
+                                EXACTFU_ONLY8   | EXACTFU
                                 EXACTFU_SS      | EXACTFU
                                 EXACTFAA        | EXACTFAA
                                 EXACTL          | EXACTL
@@ -4682,7 +4686,9 @@ S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
                        ? NOTHING                                            \
                        : ( EXACT == (X) || EXACT_ONLY8 == (X) )             \
                          ? EXACT                                            \
-                         : ( EXACTFU == (X) || EXACTFU_SS == (X) )          \
+                         : (     EXACTFU == (X)                             \
+                              || EXACTFU_ONLY8 == (X)                       \
+                              || EXACTFU_SS == (X) )                        \
                            ? EXACTFU                                        \
                            : ( EXACTFAA == (X) )                             \
                              ? EXACTFAA                                      \
@@ -13759,6 +13765,8 @@ S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
              * target string is (also) in UTF-8 */
             bool requires_utf8_target = FALSE;
 
+            bool has_micro_sign = FALSE;
+
             /* Allocate an EXACT node.  The node_type may change below to
              * another EXACTish node, but since the size of the node doesn't
              * change, it works */
@@ -14222,6 +14230,9 @@ S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
 
                             if (ender > 255)  {
                                 requires_utf8_target = TRUE;
+                                if (UNLIKELY(ender == GREEK_SMALL_LETTER_MU)) {
+                                    has_micro_sign = TRUE;
+                                }
                             }
                         }
                     }
@@ -14264,6 +14275,10 @@ S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
                         }
 #endif
 
+                        else if (UNLIKELY(ender == MICRO_SIGN)) {
+                            has_micro_sign = TRUE;
+                        }
+
                         /* Even when folding, we store just the input
                          * character, as we have an array that finds its fold
                          * quickly */
@@ -14481,6 +14496,16 @@ S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
                     else if (node_type == EXACTF) {
                         RExC_seen_d_op = TRUE;
                     }
+
+                    /* The micro sign is the only below 256 character that
+                     * folds to above 255 */
+                    if (   OP(REGNODE_p(ret)) == EXACTFU
+                        && requires_utf8_target
+                        && LIKELY(! has_micro_sign))
+                    {
+                        OP(REGNODE_p(ret)) = EXACTFU_ONLY8;
+                    }
+
                 }
 
                 alloc_maybe_populate_EXACT(pRExC_state, ret, flagp, len,
@@ -19252,6 +19277,7 @@ S_regtail_study(pTHX_ RExC_state_t *pRExC_state, regnode_offset p,
                 case EXACTFAA_NO_TRIE:
                 case EXACTFAA:
                 case EXACTFU:
+                case EXACTFU_ONLY8:
                 case EXACTFLU8:
                 case EXACTFU_SS:
                 case EXACTFL:
index 08fea13..f545078 100644 (file)
@@ -112,6 +112,9 @@ EXACTFLU8   EXACT,      str   ; Rare circumstances: like EXACTFU, but is under /
 EXACTFAA_NO_TRIE  EXACT, str     ; Match this string (which is not trie-able; not guaranteed to be folded) using /iaa rules (w/len).
 
 EXACT_ONLY8 EXACT,      str       ; Like EXACT, but no strings that aren't in UTF-8 can match
+EXACTFU_ONLY8 EXACT,    str       ; Like EXACTFU, but no strings that aren't in UTF-8 can match
+# One could add EXACTFAA8 and and something that has the same effect for /l,
+# but these would be extremely uncommon
 
 #*Do nothing types
 
index a69fdea..eb24567 100644 (file)
--- a/regexec.c
+++ b/regexec.c
@@ -2292,6 +2292,14 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s,
                                              | FOLDEQ_S2_FOLDS_SANE;
             goto do_exactf_utf8;
 
+    case EXACTFU_ONLY8:
+        if (! utf8_target) {
+            break;
+        }
+        assert(is_utf8_pat);
+        utf8_fold_flags = FOLDEQ_S2_ALREADY_FOLDED;
+        goto do_exactf_utf8;
+
     case EXACTFU:
         if (is_utf8_pat || utf8_target) {
             utf8_fold_flags = is_utf8_pat ? FOLDEQ_S2_ALREADY_FOLDED : 0;
@@ -6361,6 +6369,14 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog)
            fold_array = PL_fold_latin1;
            goto do_exactf;
 
+        case EXACTFU_ONLY8:      /* /abc/iu with something in /abc/ > 255 */
+            if (! utf8_target) {
+                sayNO;
+            }
+           assert(is_utf8_pat);
+           fold_utf8_flags = FOLDEQ_S1_ALREADY_FOLDED;
+           goto do_exactf;
+
        case EXACTFU_SS:         /*  /\x{df}/iu   */
        case EXACTFU:            /*  /abc/iu      */
            folder = foldEQ_latin1;
@@ -9288,6 +9304,14 @@ S_regrepeat(pTHX_ regexp *prog, char **startposp, const regnode *p,
                                     | FOLDEQ_S2_FOLDS_SANE;
         goto do_exactf;
 
+    case EXACTFU_ONLY8:
+        if (! utf8_target) {
+            break;
+        }
+       assert(reginfo->is_utf8_pat);
+       utf8_flags = FOLDEQ_S2_ALREADY_FOLDED;
+        goto do_exactf;
+
     case EXACTFU_SS:
     case EXACTFU:
        utf8_flags = reginfo->is_utf8_pat ? FOLDEQ_S2_ALREADY_FOLDED : 0;
index 53f8d9c..9cddfd3 100644 (file)
@@ -6,8 +6,8 @@
 
 /* Regops and State definitions */
 
-#define REGNODE_MAX            100
-#define REGMATCH_STATE_MAX     140
+#define REGNODE_MAX            101
+#define REGMATCH_STATE_MAX     141
 
 #define        END                     0       /* 0000 End of program. */
 #define        SUCCEED                 1       /* 0x01 Return from a subroutine, basically. */
 #define        EXACTFLU8               43      /* 0x2b Rare circumstances: like EXACTFU, but is under /l, UTF-8, folded, and everything in it is above 255. */
 #define        EXACTFAA_NO_TRIE        44      /* 0x2c Match this string (which is not trie-able; not guaranteed to be folded) using /iaa rules (w/len). */
 #define        EXACT_ONLY8             45      /* 0x2d Like EXACT, but no strings that aren't in UTF-8 can match */
-#define        NOTHING                 46      /* 0x2e Match empty string. */
-#define        TAIL                    47      /* 0x2f Match empty string. Can jump here from outside. */
-#define        STAR                    48      /* 0x30 Match this (simple) thing 0 or more times. */
-#define        PLUS                    49      /* 0x31 Match this (simple) thing 1 or more times. */
-#define        CURLY                   50      /* 0x32 Match this simple thing {n,m} times. */
-#define        CURLYN                  51      /* 0x33 Capture next-after-this simple thing */
-#define        CURLYM                  52      /* 0x34 Capture this medium-complex thing {n,m} times. */
-#define        CURLYX                  53      /* 0x35 Match this complex thing {n,m} times. */
-#define        WHILEM                  54      /* 0x36 Do curly processing and see if rest matches. */
-#define        OPEN                    55      /* 0x37 Mark this point in input as start of #n. */
-#define        CLOSE                   56      /* 0x38 Close corresponding OPEN of #n. */
-#define        SROPEN                  57      /* 0x39 Same as OPEN, but for script run */
-#define        SRCLOSE                 58      /* 0x3a Close preceding SROPEN */
-#define        REF                     59      /* 0x3b Match some already matched string */
-#define        REFF                    60      /* 0x3c Match already matched string, folded using native charset rules for non-utf8 */
-#define        REFFL                   61      /* 0x3d Match already matched string, folded in loc. */
-#define        REFFU                   62      /* 0x3e Match already matched string, folded using unicode rules for non-utf8 */
-#define        REFFA                   63      /* 0x3f Match already matched string, folded using unicode rules for non-utf8, no mixing ASCII, non-ASCII */
-#define        NREF                    64      /* 0x40 Match some already matched string */
-#define        NREFF                   65      /* 0x41 Match already matched string, folded using native charset rules for non-utf8 */
-#define        NREFFL                  66      /* 0x42 Match already matched string, folded in loc. */
-#define        NREFFU                  67      /* 0x43 Match already matched string, folded using unicode rules for non-utf8 */
-#define        NREFFA                  68      /* 0x44 Match already matched string, folded using unicode rules for non-utf8, no mixing ASCII, non-ASCII */
-#define        LONGJMP                 69      /* 0x45 Jump far away. */
-#define        BRANCHJ                 70      /* 0x46 BRANCH with long offset. */
-#define        IFMATCH                 71      /* 0x47 Succeeds if the following matches. */
-#define        UNLESSM                 72      /* 0x48 Fails if the following matches. */
-#define        SUSPEND                 73      /* 0x49 "Independent" sub-RE. */
-#define        IFTHEN                  74      /* 0x4a Switch, should be preceded by switcher. */
-#define        GROUPP                  75      /* 0x4b Whether the group matched. */
-#define        EVAL                    76      /* 0x4c Execute some Perl code. */
-#define        MINMOD                  77      /* 0x4d Next operator is not greedy. */
-#define        LOGICAL                 78      /* 0x4e Next opcode should set the flag only. */
-#define        RENUM                   79      /* 0x4f Group with independently numbered parens. */
-#define        TRIE                    80      /* 0x50 Match many EXACT(F[ALU]?)? at once. flags==type */
-#define        TRIEC                   81      /* 0x51 Same as TRIE, but with embedded charclass data */
-#define        AHOCORASICK             82      /* 0x52 Aho Corasick stclass. flags==type */
-#define        AHOCORASICKC            83      /* 0x53 Same as AHOCORASICK, but with embedded charclass data */
-#define        GOSUB                   84      /* 0x54 recurse to paren arg1 at (signed) ofs arg2 */
-#define        NGROUPP                 85      /* 0x55 Whether the group matched. */
-#define        INSUBP                  86      /* 0x56 Whether we are in a specific recurse. */
-#define        DEFINEP                 87      /* 0x57 Never execute directly. */
-#define        ENDLIKE                 88      /* 0x58 Used only for the type field of verbs */
-#define        OPFAIL                  89      /* 0x59 Same as (?!), but with verb arg */
-#define        ACCEPT                  90      /* 0x5a Accepts the current matched string, with verbar */
-#define        VERB                    91      /* 0x5b Used only for the type field of verbs */
-#define        PRUNE                   92      /* 0x5c Pattern fails at this startpoint if no-backtracking through this */
-#define        MARKPOINT               93      /* 0x5d Push the current location for rollback by cut. */
-#define        SKIP                    94      /* 0x5e On failure skip forward (to the mark) before retrying */
-#define        COMMIT                  95      /* 0x5f Pattern fails outright if backtracking through this */
-#define        CUTGROUP                96      /* 0x60 On failure go to the next alternation in the group */
-#define        KEEPS                   97      /* 0x61 $& begins here. */
-#define        LNBREAK                 98      /* 0x62 generic newline pattern */
-#define        OPTIMIZED               99      /* 0x63 Placeholder for dump. */
-#define        PSEUDO                  100     /* 0x64 Pseudo opcode for internal use. */
+#define        EXACTFU_ONLY8           46      /* 0x2e Like EXACTFU, but no strings that aren't in UTF-8 can match */
+#define        NOTHING                 47      /* 0x2f Match empty string. */
+#define        TAIL                    48      /* 0x30 Match empty string. Can jump here from outside. */
+#define        STAR                    49      /* 0x31 Match this (simple) thing 0 or more times. */
+#define        PLUS                    50      /* 0x32 Match this (simple) thing 1 or more times. */
+#define        CURLY                   51      /* 0x33 Match this simple thing {n,m} times. */
+#define        CURLYN                  52      /* 0x34 Capture next-after-this simple thing */
+#define        CURLYM                  53      /* 0x35 Capture this medium-complex thing {n,m} times. */
+#define        CURLYX                  54      /* 0x36 Match this complex thing {n,m} times. */
+#define        WHILEM                  55      /* 0x37 Do curly processing and see if rest matches. */
+#define        OPEN                    56      /* 0x38 Mark this point in input as start of #n. */
+#define        CLOSE                   57      /* 0x39 Close corresponding OPEN of #n. */
+#define        SROPEN                  58      /* 0x3a Same as OPEN, but for script run */
+#define        SRCLOSE                 59      /* 0x3b Close preceding SROPEN */
+#define        REF                     60      /* 0x3c Match some already matched string */
+#define        REFF                    61      /* 0x3d Match already matched string, folded using native charset rules for non-utf8 */
+#define        REFFL                   62      /* 0x3e Match already matched string, folded in loc. */
+#define        REFFU                   63      /* 0x3f Match already matched string, folded using unicode rules for non-utf8 */
+#define        REFFA                   64      /* 0x40 Match already matched string, folded using unicode rules for non-utf8, no mixing ASCII, non-ASCII */
+#define        NREF                    65      /* 0x41 Match some already matched string */
+#define        NREFF                   66      /* 0x42 Match already matched string, folded using native charset rules for non-utf8 */
+#define        NREFFL                  67      /* 0x43 Match already matched string, folded in loc. */
+#define        NREFFU                  68      /* 0x44 Match already matched string, folded using unicode rules for non-utf8 */
+#define        NREFFA                  69      /* 0x45 Match already matched string, folded using unicode rules for non-utf8, no mixing ASCII, non-ASCII */
+#define        LONGJMP                 70      /* 0x46 Jump far away. */
+#define        BRANCHJ                 71      /* 0x47 BRANCH with long offset. */
+#define        IFMATCH                 72      /* 0x48 Succeeds if the following matches. */
+#define        UNLESSM                 73      /* 0x49 Fails if the following matches. */
+#define        SUSPEND                 74      /* 0x4a "Independent" sub-RE. */
+#define        IFTHEN                  75      /* 0x4b Switch, should be preceded by switcher. */
+#define        GROUPP                  76      /* 0x4c Whether the group matched. */
+#define        EVAL                    77      /* 0x4d Execute some Perl code. */
+#define        MINMOD                  78      /* 0x4e Next operator is not greedy. */
+#define        LOGICAL                 79      /* 0x4f Next opcode should set the flag only. */
+#define        RENUM                   80      /* 0x50 Group with independently numbered parens. */
+#define        TRIE                    81      /* 0x51 Match many EXACT(F[ALU]?)? at once. flags==type */
+#define        TRIEC                   82      /* 0x52 Same as TRIE, but with embedded charclass data */
+#define        AHOCORASICK             83      /* 0x53 Aho Corasick stclass. flags==type */
+#define        AHOCORASICKC            84      /* 0x54 Same as AHOCORASICK, but with embedded charclass data */
+#define        GOSUB                   85      /* 0x55 recurse to paren arg1 at (signed) ofs arg2 */
+#define        NGROUPP                 86      /* 0x56 Whether the group matched. */
+#define        INSUBP                  87      /* 0x57 Whether we are in a specific recurse. */
+#define        DEFINEP                 88      /* 0x58 Never execute directly. */
+#define        ENDLIKE                 89      /* 0x59 Used only for the type field of verbs */
+#define        OPFAIL                  90      /* 0x5a Same as (?!), but with verb arg */
+#define        ACCEPT                  91      /* 0x5b Accepts the current matched string, with verbar */
+#define        VERB                    92      /* 0x5c Used only for the type field of verbs */
+#define        PRUNE                   93      /* 0x5d Pattern fails at this startpoint if no-backtracking through this */
+#define        MARKPOINT               94      /* 0x5e Push the current location for rollback by cut. */
+#define        SKIP                    95      /* 0x5f On failure skip forward (to the mark) before retrying */
+#define        COMMIT                  96      /* 0x60 Pattern fails outright if backtracking through this */
+#define        CUTGROUP                97      /* 0x61 On failure go to the next alternation in the group */
+#define        KEEPS                   98      /* 0x62 $& begins here. */
+#define        LNBREAK                 99      /* 0x63 generic newline pattern */
+#define        OPTIMIZED               100     /* 0x64 Placeholder for dump. */
+#define        PSEUDO                  101     /* 0x65 Pseudo opcode for internal use. */
        /* ------------ States ------------- */
 #define        TRIE_next               (REGNODE_MAX + 1)       /* state for TRIE */
 #define        TRIE_next_fail          (REGNODE_MAX + 2)       /* state for TRIE */
@@ -206,6 +207,7 @@ EXTCONST U8 PL_regkind[] = {
        EXACT,          /* EXACTFLU8              */
        EXACT,          /* EXACTFAA_NO_TRIE       */
        EXACT,          /* EXACT_ONLY8            */
+       EXACT,          /* EXACTFU_ONLY8          */
        NOTHING,        /* NOTHING                */
        NOTHING,        /* TAIL                   */
        STAR,           /* STAR                   */
@@ -356,6 +358,7 @@ static const U8 regarglen[] = {
        0,                                      /* EXACTFLU8    */
        0,                                      /* EXACTFAA_NO_TRIE */
        0,                                      /* EXACT_ONLY8  */
+       0,                                      /* EXACTFU_ONLY8 */
        0,                                      /* NOTHING      */
        0,                                      /* TAIL         */
        0,                                      /* STAR         */
@@ -462,6 +465,7 @@ static const char reg_off_by_arg[] = {
        0,      /* EXACTFLU8    */
        0,      /* EXACTFAA_NO_TRIE */
        0,      /* EXACT_ONLY8  */
+       0,      /* EXACTFU_ONLY8 */
        0,      /* NOTHING      */
        0,      /* TAIL         */
        0,      /* STAR         */
@@ -574,61 +578,62 @@ EXTCONST char * const PL_reg_name[] = {
        "EXACTFLU8",                    /* 0x2b */
        "EXACTFAA_NO_TRIE",             /* 0x2c */
        "EXACT_ONLY8",                  /* 0x2d */
-       "NOTHING",                      /* 0x2e */
-       "TAIL",                         /* 0x2f */
-       "STAR",                         /* 0x30 */
-       "PLUS",                         /* 0x31 */
-       "CURLY",                        /* 0x32 */
-       "CURLYN",                       /* 0x33 */
-       "CURLYM",                       /* 0x34 */
-       "CURLYX",                       /* 0x35 */
-       "WHILEM",                       /* 0x36 */
-       "OPEN",                         /* 0x37 */
-       "CLOSE",                        /* 0x38 */
-       "SROPEN",                       /* 0x39 */
-       "SRCLOSE",                      /* 0x3a */
-       "REF",                          /* 0x3b */
-       "REFF",                         /* 0x3c */
-       "REFFL",                        /* 0x3d */
-       "REFFU",                        /* 0x3e */
-       "REFFA",                        /* 0x3f */
-       "NREF",                         /* 0x40 */
-       "NREFF",                        /* 0x41 */
-       "NREFFL",                       /* 0x42 */
-       "NREFFU",                       /* 0x43 */
-       "NREFFA",                       /* 0x44 */
-       "LONGJMP",                      /* 0x45 */
-       "BRANCHJ",                      /* 0x46 */
-       "IFMATCH",                      /* 0x47 */
-       "UNLESSM",                      /* 0x48 */
-       "SUSPEND",                      /* 0x49 */
-       "IFTHEN",                       /* 0x4a */
-       "GROUPP",                       /* 0x4b */
-       "EVAL",                         /* 0x4c */
-       "MINMOD",                       /* 0x4d */
-       "LOGICAL",                      /* 0x4e */
-       "RENUM",                        /* 0x4f */
-       "TRIE",                         /* 0x50 */
-       "TRIEC",                        /* 0x51 */
-       "AHOCORASICK",                  /* 0x52 */
-       "AHOCORASICKC",                 /* 0x53 */
-       "GOSUB",                        /* 0x54 */
-       "NGROUPP",                      /* 0x55 */
-       "INSUBP",                       /* 0x56 */
-       "DEFINEP",                      /* 0x57 */
-       "ENDLIKE",                      /* 0x58 */
-       "OPFAIL",                       /* 0x59 */
-       "ACCEPT",                       /* 0x5a */
-       "VERB",                         /* 0x5b */
-       "PRUNE",                        /* 0x5c */
-       "MARKPOINT",                    /* 0x5d */
-       "SKIP",                         /* 0x5e */
-       "COMMIT",                       /* 0x5f */
-       "CUTGROUP",                     /* 0x60 */
-       "KEEPS",                        /* 0x61 */
-       "LNBREAK",                      /* 0x62 */
-       "OPTIMIZED",                    /* 0x63 */
-       "PSEUDO",                       /* 0x64 */
+       "EXACTFU_ONLY8",                /* 0x2e */
+       "NOTHING",                      /* 0x2f */
+       "TAIL",                         /* 0x30 */
+       "STAR",                         /* 0x31 */
+       "PLUS",                         /* 0x32 */
+       "CURLY",                        /* 0x33 */
+       "CURLYN",                       /* 0x34 */
+       "CURLYM",                       /* 0x35 */
+       "CURLYX",                       /* 0x36 */
+       "WHILEM",                       /* 0x37 */
+       "OPEN",                         /* 0x38 */
+       "CLOSE",                        /* 0x39 */
+       "SROPEN",                       /* 0x3a */
+       "SRCLOSE",                      /* 0x3b */
+       "REF",                          /* 0x3c */
+       "REFF",                         /* 0x3d */
+       "REFFL",                        /* 0x3e */
+       "REFFU",                        /* 0x3f */
+       "REFFA",                        /* 0x40 */
+       "NREF",                         /* 0x41 */
+       "NREFF",                        /* 0x42 */
+       "NREFFL",                       /* 0x43 */
+       "NREFFU",                       /* 0x44 */
+       "NREFFA",                       /* 0x45 */
+       "LONGJMP",                      /* 0x46 */
+       "BRANCHJ",                      /* 0x47 */
+       "IFMATCH",                      /* 0x48 */
+       "UNLESSM",                      /* 0x49 */
+       "SUSPEND",                      /* 0x4a */
+       "IFTHEN",                       /* 0x4b */
+       "GROUPP",                       /* 0x4c */
+       "EVAL",                         /* 0x4d */
+       "MINMOD",                       /* 0x4e */
+       "LOGICAL",                      /* 0x4f */
+       "RENUM",                        /* 0x50 */
+       "TRIE",                         /* 0x51 */
+       "TRIEC",                        /* 0x52 */
+       "AHOCORASICK",                  /* 0x53 */
+       "AHOCORASICKC",                 /* 0x54 */
+       "GOSUB",                        /* 0x55 */
+       "NGROUPP",                      /* 0x56 */
+       "INSUBP",                       /* 0x57 */
+       "DEFINEP",                      /* 0x58 */
+       "ENDLIKE",                      /* 0x59 */
+       "OPFAIL",                       /* 0x5a */
+       "ACCEPT",                       /* 0x5b */
+       "VERB",                         /* 0x5c */
+       "PRUNE",                        /* 0x5d */
+       "MARKPOINT",                    /* 0x5e */
+       "SKIP",                         /* 0x5f */
+       "COMMIT",                       /* 0x60 */
+       "CUTGROUP",                     /* 0x61 */
+       "KEEPS",                        /* 0x62 */
+       "LNBREAK",                      /* 0x63 */
+       "OPTIMIZED",                    /* 0x64 */
+       "PSEUDO",                       /* 0x65 */
        /* ------------ States ------------- */
        "TRIE_next",                    /* REGNODE_MAX +0x01 */
        "TRIE_next_fail",               /* REGNODE_MAX +0x02 */
@@ -763,7 +768,7 @@ EXTCONST U8 PL_varies[] __attribute__deprecated__ = {
 EXTCONST U8 PL_varies_bitmask[];
 #else
 EXTCONST U8 PL_varies_bitmask[] = {
-    0x00, 0x00, 0x00, 0x00, 0x0C, 0x00, 0x7F, 0xF8, 0x5F, 0x06, 0x00, 0x00, 0x00
+    0x00, 0x00, 0x00, 0x00, 0x0C, 0x00, 0xFE, 0xF0, 0xBF, 0x0C, 0x00, 0x00, 0x00
 };
 #endif /* DOINIT */