This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Mention http://sourceforge.net/p/predef/wiki/Home/
[perl5.git] / pod / perlhacktips.pod
CommitLineData
0061d4fa 1
04c692a8
DR
2=encoding utf8
3
4=for comment
5Consistent formatting of this file is achieved with:
6 perl ./Porting/podtidy pod/perlhacktips.pod
7
8=head1 NAME
9
10perlhacktips - Tips for Perl core C code hacking
11
12=head1 DESCRIPTION
13
14This document will help you learn the best way to go about hacking on
9b22382a 15the Perl core C code. It covers common problems, debugging, profiling,
04c692a8
DR
16and more.
17
18If you haven't read L<perlhack> and L<perlhacktut> yet, you might want
19to do that first.
20
21=head1 COMMON PROBLEMS
22
9b22382a 23Perl source plays by ANSI C89 rules: no C99 (or C++) extensions. In
04c692a8
DR
24some cases we have to take pre-ANSI requirements into consideration.
25You don't care about some particular platform having broken Perl? I
26hear there is still a strong demand for J2EE programmers.
27
28=head2 Perl environment problems
29
30=over 4
31
32=item *
33
34Not compiling with threading
35
36Compiling with threading (-Duseithreads) completely rewrites the
9b22382a 37function prototypes of Perl. You better try your changes with that.
04c692a8
DR
38Related to this is the difference between "Perl_-less" and "Perl_-ly"
39APIs, for example:
40
41 Perl_sv_setiv(aTHX_ ...);
42 sv_setiv(...);
43
44The first one explicitly passes in the context, which is needed for
9b22382a
FC
45e.g. threaded builds. The second one does that implicitly; do not get
46them mixed. If you are not passing in a aTHX_, you will need to do a
04c692a8
DR
47dTHX (or a dVAR) as the first thing in the function.
48
49See L<perlguts/"How multiple interpreters and concurrency are
50supported"> for further discussion about context.
51
52=item *
53
54Not compiling with -DDEBUGGING
55
56The DEBUGGING define exposes more code to the compiler, therefore more
9b22382a 57ways for things to go wrong. You should try it.
04c692a8
DR
58
59=item *
60
61Introducing (non-read-only) globals
62
63Do not introduce any modifiable globals, truly global or file static.
64They are bad form and complicate multithreading and other forms of
9b22382a 65concurrency. The right way is to introduce them as new interpreter
04c692a8
DR
66variables, see F<intrpvar.h> (at the very end for binary
67compatibility).
68
69Introducing read-only (const) globals is okay, as long as you verify
70with e.g. C<nm libperl.a|egrep -v ' [TURtr] '> (if your C<nm> has
9b22382a 71BSD-style output) that the data you added really is read-only. (If it
04c692a8
DR
72is, it shouldn't show up in the output of that command.)
73
74If you want to have static strings, make them constant:
75
76 static const char etc[] = "...";
77
78If you want to have arrays of constant strings, note carefully the
79right combination of C<const>s:
80
81 static const char * const yippee[] =
a63ef199 82 {"hi", "ho", "silver"};
04c692a8
DR
83
84There is a way to completely hide any modifiable globals (they are all
85moved to heap), the compilation setting
9b22382a 86C<-DPERL_GLOBAL_STRUCT_PRIVATE>. It is not normally used, but can be
04c692a8
DR
87used for testing, read more about it in L<perlguts/"Background and
88PERL_IMPLICIT_CONTEXT">.
89
90=item *
91
92Not exporting your new function
93
94Some platforms (Win32, AIX, VMS, OS/2, to name a few) require any
95function that is part of the public API (the shared Perl library) to be
9b22382a 96explicitly marked as exported. See the discussion about F<embed.pl> in
04c692a8
DR
97L<perlguts>.
98
99=item *
100
101Exporting your new function
102
103The new shiny result of either genuine new functionality or your
9b22382a 104arduous refactoring is now ready and correctly exported. So what could
04c692a8
DR
105possibly go wrong?
106
107Maybe simply that your function did not need to be exported in the
9b22382a 108first place. Perl has a long and not so glorious history of exporting
04c692a8
DR
109functions that it should not have.
110
111If the function is used only inside one source code file, make it
9b22382a 112static. See the discussion about F<embed.pl> in L<perlguts>.
04c692a8
DR
113
114If the function is used across several files, but intended only for
115Perl's internal use (and this should be the common case), do not export
9b22382a 116it to the public API. See the discussion about F<embed.pl> in
04c692a8
DR
117L<perlguts>.
118
119=back
120
121=head2 Portability problems
122
123The following are common causes of compilation and/or execution
9b22382a
FC
124failures, not common to Perl as such. The C FAQ is good bedtime
125reading. Please test your changes with as many C compilers and
04c692a8
DR
126platforms as possible; we will, anyway, and it's nice to save oneself
127from public embarrassment.
128
129If using gcc, you can add the C<-std=c89> option which will hopefully
9b22382a 130catch most of these unportabilities. (However it might also catch
04c692a8
DR
131incompatibilities in your system's header files.)
132
133Use the Configure C<-Dgccansipedantic> flag to enable the gcc C<-ansi
134-pedantic> flags which enforce stricter ANSI rules.
135
136If using the C<gcc -Wall> note that not all the possible warnings (like
137C<-Wunitialized>) are given unless you also compile with C<-O>.
138
139Note that if using gcc, starting from Perl 5.9.5 the Perl core source
140code files (the ones at the top level of the source code distribution,
141but not e.g. the extensions under ext/) are automatically compiled with
142as many as possible of the C<-std=c89>, C<-ansi>, C<-pedantic>, and a
143selection of C<-W> flags (see cflags.SH).
144
145Also study L<perlport> carefully to avoid any bad assumptions about the
eb9df707 146operating system, filesystems, character set, and so forth.
04c692a8
DR
147
148You may once in a while try a "make microperl" to see whether we can
9b22382a 149still compile Perl with just the bare minimum of interfaces. (See
04c692a8
DR
150README.micro.)
151
152Do not assume an operating system indicates a certain compiler.
153
154=over 4
155
156=item *
157
158Casting pointers to integers or casting integers to pointers
159
160 void castaway(U8* p)
161 {
162 IV i = p;
163
164or
165
166 void castaway(U8* p)
167 {
168 IV i = (IV)p;
169
9b22382a
FC
170Both are bad, and broken, and unportable. Use the PTR2IV() macro that
171does it right. (Likewise, there are PTR2UV(), PTR2NV(), INT2PTR(), and
04c692a8
DR
172NUM2PTR().)
173
174=item *
175
28ffebaf 176Casting between function pointers and data pointers
04c692a8
DR
177
178Technically speaking casting between function pointers and data
179pointers is unportable and undefined, but practically speaking it seems
180to work, but you should use the FPTR2DPTR() and DPTR2FPTR() macros.
181Sometimes you can also play games with unions.
182
183=item *
184
185Assuming sizeof(int) == sizeof(long)
186
187There are platforms where longs are 64 bits, and platforms where ints
188are 64 bits, and while we are out to shock you, even platforms where
9b22382a 189shorts are 64 bits. This is all legal according to the C standard. (In
04c692a8
DR
190other words, "long long" is not a portable way to specify 64 bits, and
191"long long" is not even guaranteed to be any wider than "long".)
192
193Instead, use the definitions IV, UV, IVSIZE, I32SIZE, and so forth.
194Avoid things like I32 because they are B<not> guaranteed to be
195I<exactly> 32 bits, they are I<at least> 32 bits, nor are they
9b22382a 196guaranteed to be B<int> or B<long>. If you really explicitly need
04c692a8
DR
19764-bit variables, use I64 and U64, but only if guarded by HAS_QUAD.
198
199=item *
200
201Assuming one can dereference any type of pointer for any type of data
202
203 char *p = ...;
204 long pony = *p; /* BAD */
205
206Many platforms, quite rightly so, will give you a core dump instead of
768312ab 207a pony if the p happens not to be correctly aligned.
04c692a8
DR
208
209=item *
210
211Lvalue casts
212
213 (int)*p = ...; /* BAD */
214
9b22382a 215Simply not portable. Get your lvalue to be of the right type, or maybe
04c692a8
DR
216use temporary variables, or dirty tricks with unions.
217
218=item *
219
220Assume B<anything> about structs (especially the ones you don't
221control, like the ones coming from the system headers)
222
223=over 8
224
225=item *
226
227That a certain field exists in a struct
228
229=item *
230
231That no other fields exist besides the ones you know of
232
233=item *
234
235That a field is of certain signedness, sizeof, or type
236
237=item *
238
239That the fields are in a certain order
240
241=over 8
242
243=item *
244
245While C guarantees the ordering specified in the struct definition,
246between different platforms the definitions might differ
247
248=back
249
250=item *
251
252That the sizeof(struct) or the alignments are the same everywhere
253
254=over 8
255
256=item *
257
258There might be padding bytes between the fields to align the fields -
259the bytes can be anything
260
261=item *
262
263Structs are required to be aligned to the maximum alignment required by
264the fields - which for native types is for usually equivalent to
265sizeof() of the field
266
267=back
268
269=back
270
271=item *
272
273Assuming the character set is ASCIIish
274
9b22382a 275Perl can compile and run under EBCDIC platforms. See L<perlebcdic>.
04c692a8
DR
276This is transparent for the most part, but because the character sets
277differ, you shouldn't use numeric (decimal, octal, nor hex) constants
eb9df707
KW
278to refer to characters. You can safely say C<'A'>, but not C<0x41>.
279You can safely say C<'\n'>, but not C<\012>. However, you can use
280macros defined in F<utf8.h> to specify any code point portably.
281C<LATIN1_TO_NATIVE(0xDF)> is going to be the code point that means
282LATIN SMALL LETTER SHARP S on whatever platform you are running on (on
283ASCII platforms it compiles without adding any extra code, so there is
284zero performance hit on those). The acceptable inputs to
285C<LATIN1_TO_NATIVE> are from C<0x00> through C<0xFF>. If your input
286isn't guaranteed to be in that range, use C<UNICODE_TO_NATIVE> instead.
287C<NATIVE_TO_LATIN1> and C<NATIVE_TO_UNICODE> translate the opposite
288direction.
289
290If you need the string representation of a character that doesn't have a
291mnemonic name in C, you should add it to the list in
c22aa07d 292F<regen/unicode_constants.pl>, and have Perl create C<#define>'s for you,
eb6d698b 293based on the current platform.
04c692a8 294
eb9df707
KW
295Note that the C<isI<FOO>> and C<toI<FOO>> macros in F<handy.h> work
296properly on native code points and strings.
297
04c692a8 298Also, the range 'A' - 'Z' in ASCII is an unbroken sequence of 26 upper
9b22382a
FC
299case alphabetic characters. That is not true in EBCDIC. Nor for 'a' to
300'z'. But '0' - '9' is an unbroken range in both systems. Don't assume
c22aa07d 301anything about other ranges. (Note that special handling of ranges in
f4240379 302regular expression patterns and transliterations makes it appear to Perl
c22aa07d 303code that the aforementioned ranges are all unbroken.)
04c692a8
DR
304
305Many of the comments in the existing code ignore the possibility of
9b22382a 306EBCDIC, and may be wrong therefore, even if the code works. This is
04c692a8
DR
307actually a tribute to the successful transparent insertion of being
308able to handle EBCDIC without having to change pre-existing code.
309
310UTF-8 and UTF-EBCDIC are two different encodings used to represent
9b22382a 311Unicode code points as sequences of bytes. Macros with the same names
eb9df707 312(but different definitions) in F<utf8.h> and F<utfebcdic.h> are used to
04c692a8
DR
313allow the calling code to think that there is only one such encoding.
314This is almost always referred to as C<utf8>, but it means the EBCDIC
9b22382a 315version as well. Again, comments in the code may well be wrong even if
eb9df707 316the code itself is right. For example, the concept of UTF-8 C<invariant
9b22382a
FC
317characters> differs between ASCII and EBCDIC. On ASCII platforms, only
318characters that do not have the high-order bit set (i.e. whose ordinals
04c692a8
DR
319are strict ASCII, 0 - 127) are invariant, and the documentation and
320comments in the code may assume that, often referring to something
9b22382a 321like, say, C<hibit>. The situation differs and is not so simple on
04c692a8
DR
322EBCDIC machines, but as long as the code itself uses the
323C<NATIVE_IS_INVARIANT()> macro appropriately, it works, even if the
324comments are wrong.
325
257844b9
KW
326As noted in L<perlhack/TESTING>, when writing test scripts, the file
327F<t/charset_tools.pl> contains some helpful functions for writing tests
328valid on both ASCII and EBCDIC platforms. Sometimes, though, a test
329can't use a function and it's inconvenient to have different test
330versions depending on the platform. There are 20 code points that are
331the same in all 4 character sets currently recognized by Perl (the 3
332EBCDIC code pages plus ISO 8859-1 (ASCII/Latin1)). These can be used in
333such tests, though there is a small possibility that Perl will become
334available in yet another character set, breaking your test. All but one
335of these code points are C0 control characters. The most significant
336controls that are the same are C<\0>, C<\r>, and C<\N{VT}> (also
337specifiable as C<\cK>, C<\x0B>, C<\N{U+0B}>, or C<\013>). The single
338non-control is U+00B6 PILCROW SIGN. The controls that are the same have
339the same bit pattern in all 4 character sets, regardless of the UTF8ness
340of the string containing them. The bit pattern for U+B6 is the same in
341all 4 for non-UTF8 strings, but differs in each when its containing
342string is UTF-8 encoded. The only other code points that have some sort
343of sameness across all 4 character sets are the pair 0xDC and 0xFC.
344Together these represent upper- and lowercase LATIN LETTER U WITH
345DIAERESIS, but which is upper and which is lower may be reversed: 0xDC
346is the capital in Latin1 and 0xFC is the small letter, while 0xFC is the
347capital in EBCDIC and 0xDC is the small one. This factoid may be
348exploited in writing case insensitive tests that are the same across all
3494 character sets.
350
04c692a8
DR
351=item *
352
353Assuming the character set is just ASCII
354
9b22382a 355ASCII is a 7 bit encoding, but bytes have 8 bits in them. The 128 extra
04c692a8
DR
356characters have different meanings depending on the locale. Absent a
357locale, currently these extra characters are generally considered to be
eb9df707
KW
358unassigned, and this has presented some problems. This has being
359changed starting in 5.12 so that these characters can be considered to
360be Latin-1 (ISO-8859-1).
04c692a8
DR
361
362=item *
363
364Mixing #define and #ifdef
365
366 #define BURGLE(x) ... \
367 #ifdef BURGLE_OLD_STYLE /* BAD */
368 ... do it the old way ... \
369 #else
370 ... do it the new way ... \
371 #endif
372
9b22382a 373You cannot portably "stack" cpp directives. For example in the above
04c692a8
DR
374you need two separate BURGLE() #defines, one for each #ifdef branch.
375
376=item *
377
378Adding non-comment stuff after #endif or #else
379
380 #ifdef SNOSH
381 ...
382 #else !SNOSH /* BAD */
383 ...
384 #endif SNOSH /* BAD */
385
386The #endif and #else cannot portably have anything non-comment after
9b22382a 387them. If you want to document what is going (which is a good idea
04c692a8
DR
388especially if the branches are long), use (C) comments:
389
390 #ifdef SNOSH
391 ...
392 #else /* !SNOSH */
393 ...
394 #endif /* SNOSH */
395
396The gcc option C<-Wendif-labels> warns about the bad variant (by
397default on starting from Perl 5.9.4).
398
399=item *
400
401Having a comma after the last element of an enum list
402
403 enum color {
404 CERULEAN,
405 CHARTREUSE,
406 CINNABAR, /* BAD */
407 };
408
9b22382a 409is not portable. Leave out the last comma.
04c692a8
DR
410
411Also note that whether enums are implicitly morphable to ints varies
412between compilers, you might need to (int).
413
414=item *
415
416Using //-comments
417
418 // This function bamfoodles the zorklator. /* BAD */
419
9b22382a 420That is C99 or C++. Perl is C89. Using the //-comments is silently
04c692a8
DR
421allowed by many C compilers but cranking up the ANSI C89 strictness
422(which we like to do) causes the compilation to fail.
423
424=item *
425
426Mixing declarations and code
427
428 void zorklator()
429 {
430 int n = 3;
431 set_zorkmids(n); /* BAD */
432 int q = 4;
433
9b22382a 434That is C99 or C++. Some C compilers allow that, but you shouldn't.
04c692a8
DR
435
436The gcc option C<-Wdeclaration-after-statements> scans for such
437problems (by default on starting from Perl 5.9.4).
438
439=item *
440
441Introducing variables inside for()
442
443 for(int i = ...; ...; ...) { /* BAD */
444
9b22382a 445That is C99 or C++. While it would indeed be awfully nice to have that
04c692a8
DR
446also in C89, to limit the scope of the loop variable, alas, we cannot.
447
448=item *
449
450Mixing signed char pointers with unsigned char pointers
451
452 int foo(char *s) { ... }
453 ...
454 unsigned char *t = ...; /* Or U8* t = ... */
455 foo(t); /* BAD */
456
457While this is legal practice, it is certainly dubious, and downright
458fatal in at least one platform: for example VMS cc considers this a
9b22382a 459fatal error. One cause for people often making this mistake is that a
04c692a8
DR
460"naked char" and therefore dereferencing a "naked char pointer" have an
461undefined signedness: it depends on the compiler and the flags of the
462compiler and the underlying platform whether the result is signed or
9b22382a 463unsigned. For this very same reason using a 'char' as an array index is
04c692a8
DR
464bad.
465
466=item *
467
468Macros that have string constants and their arguments as substrings of
469the string constants
470
471 #define FOO(n) printf("number = %d\n", n) /* BAD */
472 FOO(10);
473
474Pre-ANSI semantics for that was equivalent to
475
476 printf("10umber = %d\10");
477
9b22382a 478which is probably not what you were expecting. Unfortunately at least
04c692a8
DR
479one reasonably common and modern C compiler does "real backward
480compatibility" here, in AIX that is what still happens even though the
481rest of the AIX compiler is very happily C89.
482
483=item *
484
485Using printf formats for non-basic C types
486
487 IV i = ...;
488 printf("i = %d\n", i); /* BAD */
489
490While this might by accident work in some platform (where IV happens to
9b22382a 491be an C<int>), in general it cannot. IV might be something larger. Even
04c692a8
DR
492worse the situation is with more specific types (defined by Perl's
493configuration step in F<config.h>):
494
495 Uid_t who = ...;
496 printf("who = %d\n", who); /* BAD */
497
498The problem here is that Uid_t might be not only not C<int>-wide but it
499might also be unsigned, in which case large uids would be printed as
500negative values.
501
502There is no simple solution to this because of printf()'s limited
503intelligence, but for many types the right format is available as with
504either 'f' or '_f' suffix, for example:
505
506 IVdf /* IV in decimal */
507 UVxf /* UV is hexadecimal */
508
509 printf("i = %"IVdf"\n", i); /* The IVdf is a string constant. */
510
511 Uid_t_f /* Uid_t in decimal */
512
513 printf("who = %"Uid_t_f"\n", who);
514
515Or you can try casting to a "wide enough" type:
516
517 printf("i = %"IVdf"\n", (IV)something_very_small_and_signed);
518
519Also remember that the C<%p> format really does require a void pointer:
520
521 U8* p = ...;
522 printf("p = %p\n", (void*)p);
523
524The gcc option C<-Wformat> scans for such problems.
525
526=item *
527
528Blindly using variadic macros
529
530gcc has had them for a while with its own syntax, and C99 brought them
9b22382a 531with a standardized syntax. Don't use the former, and use the latter
04c692a8
DR
532only if the HAS_C99_VARIADIC_MACROS is defined.
533
534=item *
535
536Blindly passing va_list
537
538Not all platforms support passing va_list to further varargs (stdarg)
9b22382a 539functions. The right thing to do is to copy the va_list using the
04c692a8
DR
540Perl_va_copy() if the NEED_VA_COPY is defined.
541
542=item *
543
544Using gcc statement expressions
545
546 val = ({...;...;...}); /* BAD */
547
9b22382a 548While a nice extension, it's not portable. The Perl code does
04c692a8
DR
549admittedly use them if available to gain some extra speed (essentially
550as a funky form of inlining), but you shouldn't.
551
552=item *
553
554Binding together several statements in a macro
555
556Use the macros STMT_START and STMT_END.
557
558 STMT_START {
559 ...
560 } STMT_END
561
562=item *
563
564Testing for operating systems or versions when should be testing for
565features
566
567 #ifdef __FOONIX__ /* BAD */
568 foo = quux();
569 #endif
570
571Unless you know with 100% certainty that quux() is only ever available
572for the "Foonix" operating system B<and> that is available B<and>
573correctly working for B<all> past, present, B<and> future versions of
9b22382a 574"Foonix", the above is very wrong. This is more correct (though still
04c692a8
DR
575not perfect, because the below is a compile-time check):
576
577 #ifdef HAS_QUUX
578 foo = quux();
579 #endif
580
581How does the HAS_QUUX become defined where it needs to be? Well, if
582Foonix happens to be Unixy enough to be able to run the Configure
583script, and Configure has been taught about detecting and testing
9b22382a 584quux(), the HAS_QUUX will be correctly defined. In other platforms, the
04c692a8
DR
585corresponding configuration step will hopefully do the same.
586
587In a pinch, if you cannot wait for Configure to be educated, or if you
588have a good hunch of where quux() might be available, you can
589temporarily try the following:
590
591 #if (defined(__FOONIX__) || defined(__BARNIX__))
592 # define HAS_QUUX
593 #endif
594
595 ...
596
597 #ifdef HAS_QUUX
598 foo = quux();
599 #endif
600
601But in any case, try to keep the features and operating systems
602separate.
603
b39b5b0c
JH
604A good resource on the predefined macros for various operating
605systems, compilers, and so forth is
606L<http://sourceforge.net/p/predef/wiki/Home/>
607
38f18a30
KW
608=item *
609
610Assuming the contents of static memory pointed to by the return values
611of Perl wrappers for C library functions doesn't change. Many C library
612functions return pointers to static storage that can be overwritten by
613subsequent calls to the same or related functions. Perl has
614light-weight wrappers for some of these functions, and which don't make
615copies of the static memory. A good example is the interface to the
616environment variables that are in effect for the program. Perl has
617C<PerlEnv_getenv> to get values from the environment. But the return is
618a pointer to static memory in the C library. If you are using the value
619to immediately test for something, that's fine, but if you save the
620value and expect it to be unchanged by later processing, you would be
621wrong, but perhaps you wouldn't know it because different C library
622implementations behave differently, and the one on the platform you're
623testing on might work for your situation. But on some platforms, a
624subsequent call to C<PerlEnv_getenv> or related function WILL overwrite
625the memory that your first call points to. This has led to some
626hard-to-debug problems. Do a L<perlapi/savepv> to make a copy, thus
627avoiding these problems. You will have to free the copy when you're
628done to avoid memory leaks. If you don't have control over when it gets
629freed, you'll need to make the copy in a mortal scalar, like so:
630
631 if ((s = PerlEnv_getenv("foo") == NULL) {
632 ... /* handle NULL case */
633 }
634 else {
635 s = SvPVX(sv_2mortal(newSVpv(s, 0)));
636 }
637
638The above example works only if C<"s"> is C<NUL>-terminated; otherwise
639you have to pass its length to C<newSVpv>.
640
04c692a8
DR
641=back
642
643=head2 Problematic System Interfaces
644
645=over 4
646
647=item *
648
9b22382a
FC
649malloc(0), realloc(0), calloc(0, 0) are non-portable. To be portable
650allocate at least one byte. (In general you should rarely need to work
04c692a8
DR
651at this low level, but instead use the various malloc wrappers.)
652
653=item *
654
9b22382a 655snprintf() - the return type is unportable. Use my_snprintf() instead.
04c692a8
DR
656
657=back
658
659=head2 Security problems
660
661Last but not least, here are various tips for safer coding.
bbc89b61 662See also L<perlclib> for libc/stdio replacements one should use.
04c692a8
DR
663
664=over 4
665
666=item *
667
668Do not use gets()
669
9b22382a 670Or we will publicly ridicule you. Seriously.
04c692a8
DR
671
672=item *
673
bbc89b61
JH
674Do not use tmpfile()
675
676Use mkstemp() instead.
677
678=item *
679
04c692a8
DR
680Do not use strcpy() or strcat() or strncpy() or strncat()
681
682Use my_strlcpy() and my_strlcat() instead: they either use the native
683implementation, or Perl's own implementation (borrowed from the public
684domain implementation of INN).
685
686=item *
687
688Do not use sprintf() or vsprintf()
689
690If you really want just plain byte strings, use my_snprintf() and
691my_vsnprintf() instead, which will try to use snprintf() and
9b22382a 692vsnprintf() if those safer APIs are available. If you want something
6bfe0388
KW
693fancier than a plain byte string, use
694L<C<Perl_form>()|perlapi/form> or SVs and
695L<C<Perl_sv_catpvf()>|perlapi/sv_catpvf>.
696
2e642750
KW
697Note that glibc C<printf()>, C<sprintf()>, etc. are buggy before glibc
698version 2.17. They won't allow a C<%.s> format with a precision to
699create a string that isn't valid UTF-8 if the current underlying locale
700of the program is UTF-8. What happens is that the C<%s> and its operand are
6bfe0388 701simply skipped without any notice.
2e642750 702L<https://sourceware.org/bugzilla/show_bug.cgi?id=6530>.
04c692a8 703
c98823ff
JH
704=item *
705
706Do not use atoi()
707
22ff3130 708Use grok_atoUV() instead. atoi() has ill-defined behavior on overflows,
c98823ff 709and cannot be used for incremental parsing. It is also affected by locale,
338aa8b0
JH
710which is bad.
711
712=item *
713
714Do not use strtol() or strtoul()
715
22ff3130 716Use grok_atoUV() instead. strtol() or strtoul() (or their IV/UV-friendly
338aa8b0
JH
717macro disguises, Strtol() and Strtoul(), or Atol() and Atoul() are
718affected by locale, which is bad.
c98823ff 719
04c692a8
DR
720=back
721
722=head1 DEBUGGING
723
724You can compile a special debugging version of Perl, which allows you
725to use the C<-D> option of Perl to tell more about what Perl is doing.
726But sometimes there is no alternative than to dive in with a debugger,
727either to see the stack trace of a core dump (very useful in a bug
728report), or trying to figure out what went wrong before the core dump
729happened, or how did we end up having wrong or unexpected results.
730
731=head2 Poking at Perl
732
733To really poke around with Perl, you'll probably want to build Perl for
734debugging, like this:
735
736 ./Configure -d -D optimize=-g
737 make
738
739C<-g> is a flag to the C compiler to have it produce debugging
740information which will allow us to step through a running program, and
741to see in which C function we are at (without the debugging information
742we might see only the numerical addresses of the functions, which is
743not very helpful).
744
745F<Configure> will also turn on the C<DEBUGGING> compilation symbol
9b22382a 746which enables all the internal debugging code in Perl. There are a
04c692a8
DR
747whole bunch of things you can debug with this: L<perlrun> lists them
748all, and the best way to find out about them is to play about with
9b22382a 749them. The most useful options are probably
04c692a8
DR
750
751 l Context (loop) stack processing
752 t Trace execution
753 o Method and overloading resolution
754 c String/numeric conversions
755
756Some of the functionality of the debugging code can be achieved using
757XS modules.
758
759 -Dr => use re 'debug'
760 -Dx => use O 'Debug'
761
762=head2 Using a source-level debugger
763
764If the debugging output of C<-D> doesn't help you, it's time to step
765through perl's execution with a source-level debugger.
766
767=over 3
768
769=item *
770
771We'll use C<gdb> for our examples here; the principles will apply to
772any debugger (many vendors call their debugger C<dbx>), but check the
773manual of the one you're using.
774
775=back
776
777To fire up the debugger, type
778
779 gdb ./perl
780
781Or if you have a core dump:
782
783 gdb ./perl core
784
785You'll want to do that in your Perl source tree so the debugger can
9b22382a 786read the source code. You should see the copyright message, followed by
04c692a8
DR
787the prompt.
788
789 (gdb)
790
791C<help> will get you into the documentation, but here are the most
792useful commands:
793
794=over 3
795
796=item * run [args]
797
798Run the program with the given arguments.
799
800=item * break function_name
801
802=item * break source.c:xxx
803
804Tells the debugger that we'll want to pause execution when we reach
805either the named function (but see L<perlguts/Internal Functions>!) or
806the given line in the named source file.
807
808=item * step
809
810Steps through the program a line at a time.
811
812=item * next
813
814Steps through the program a line at a time, without descending into
815functions.
816
817=item * continue
818
819Run until the next breakpoint.
820
821=item * finish
822
823Run until the end of the current function, then stop again.
824
825=item * 'enter'
826
827Just pressing Enter will do the most recent operation again - it's a
828blessing when stepping through miles of source code.
829
8b029fdf
MH
830=item * ptype
831
832Prints the C definition of the argument given.
833
834 (gdb) ptype PL_op
835 type = struct op {
836 OP *op_next;
86cd3a13 837 OP *op_sibparent;
8b029fdf
MH
838 OP *(*op_ppaddr)(void);
839 PADOFFSET op_targ;
840 unsigned int op_type : 9;
841 unsigned int op_opt : 1;
842 unsigned int op_slabbed : 1;
843 unsigned int op_savefree : 1;
844 unsigned int op_static : 1;
845 unsigned int op_folded : 1;
846 unsigned int op_spare : 2;
847 U8 op_flags;
848 U8 op_private;
849 } *
850
04c692a8
DR
851=item * print
852
9b22382a 853Execute the given C code and print its results. B<WARNING>: Perl makes
04c692a8 854heavy use of macros, and F<gdb> does not necessarily support macros
9b22382a 855(see later L</"gdb macro support">). You'll have to substitute them
04c692a8
DR
856yourself, or to invoke cpp on the source code files (see L</"The .i
857Targets">) So, for instance, you can't say
858
859 print SvPV_nolen(sv)
860
861but you have to say
862
863 print Perl_sv_2pv_nolen(sv)
864
865=back
866
867You may find it helpful to have a "macro dictionary", which you can
9b22382a 868produce by saying C<cpp -dM perl.c | sort>. Even then, F<cpp> won't
04c692a8
DR
869recursively apply those macros for you.
870
871=head2 gdb macro support
872
873Recent versions of F<gdb> have fairly good macro support, but in order
874to use it you'll need to compile perl with macro definitions included
9b22382a
FC
875in the debugging information. Using F<gcc> version 3.1, this means
876configuring with C<-Doptimize=-g3>. Other compilers might use a
04c692a8
DR
877different switch (if they support debugging macros at all).
878
879=head2 Dumping Perl Data Structures
880
881One way to get around this macro hell is to use the dumping functions
882in F<dump.c>; these work a little like an internal
883L<Devel::Peek|Devel::Peek>, but they also cover OPs and other
9b22382a 884structures that you can't get at from Perl. Let's take an example.
04c692a8 885We'll use the C<$a = $b + $c> we used before, but give it a bit of
9b22382a 886context: C<$b = "6XXXX"; $c = 2.3;>. Where's a good place to stop and
04c692a8
DR
887poke around?
888
889What about C<pp_add>, the function we examined earlier to implement the
890C<+> operator:
891
892 (gdb) break Perl_pp_add
893 Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.
894
895Notice we use C<Perl_pp_add> and not C<pp_add> - see
9b22382a 896L<perlguts/Internal Functions>. With the breakpoint in place, we can
04c692a8
DR
897run our program:
898
899 (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'
900
901Lots of junk will go past as gdb reads in the relevant source files and
902libraries, and then:
903
904 Breakpoint 1, Perl_pp_add () at pp_hot.c:309
905 309 dSP; dATARGET; tryAMAGICbin(add,opASSIGN);
906 (gdb) step
907 311 dPOPTOPnnrl_ul;
908 (gdb)
909
910We looked at this bit of code before, and we said that
911C<dPOPTOPnnrl_ul> arranges for two C<NV>s to be placed into C<left> and
912C<right> - let's slightly expand it:
913
914 #define dPOPTOPnnrl_ul NV right = POPn; \
915 SV *leftsv = TOPs; \
916 NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0
917
918C<POPn> takes the SV from the top of the stack and obtains its NV
919either directly (if C<SvNOK> is set) or by calling the C<sv_2nv>
9b22382a
FC
920function. C<TOPs> takes the next SV from the top of the stack - yes,
921C<POPn> uses C<TOPs> - but doesn't remove it. We then use C<SvNV> to
04c692a8
DR
922get the NV from C<leftsv> in the same way as before - yes, C<POPn> uses
923C<SvNV>.
924
925Since we don't have an NV for C<$b>, we'll have to use C<sv_2nv> to
9b22382a 926convert it. If we step again, we'll find ourselves there:
04c692a8 927
8b029fdf 928 (gdb) step
04c692a8
DR
929 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
930 1669 if (!sv)
931 (gdb)
932
933We can now use C<Perl_sv_dump> to investigate the SV:
934
8b029fdf 935 (gdb) print Perl_sv_dump(sv)
04c692a8
DR
936 SV = PV(0xa057cc0) at 0xa0675d0
937 REFCNT = 1
938 FLAGS = (POK,pPOK)
939 PV = 0xa06a510 "6XXXX"\0
940 CUR = 5
941 LEN = 6
942 $1 = void
943
944We know we're going to get C<6> from this, so let's finish the
945subroutine:
946
947 (gdb) finish
948 Run till exit from #0 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
949 0x462669 in Perl_pp_add () at pp_hot.c:311
950 311 dPOPTOPnnrl_ul;
951
952We can also dump out this op: the current op is always stored in
9b22382a 953C<PL_op>, and we can dump it with C<Perl_op_dump>. This'll give us
04c692a8
DR
954similar output to L<B::Debug|B::Debug>.
955
8b029fdf 956 (gdb) print Perl_op_dump(PL_op)
04c692a8
DR
957 {
958 13 TYPE = add ===> 14
959 TARG = 1
960 FLAGS = (SCALAR,KIDS)
961 {
962 TYPE = null ===> (12)
963 (was rv2sv)
964 FLAGS = (SCALAR,KIDS)
965 {
966 11 TYPE = gvsv ===> 12
967 FLAGS = (SCALAR)
968 GV = main::b
969 }
970 }
971
972# finish this later #
973
8b029fdf
MH
974=head2 Using gdb to look at specific parts of a program
975
73013070
SF
976With the example above, you knew to look for C<Perl_pp_add>, but what if
977there were multiple calls to it all over the place, or you didn't know what
8b029fdf
MH
978the op was you were looking for?
979
73013070 980One way to do this is to inject a rare call somewhere near what you're looking
9b22382a 981for. For example, you could add C<study> before your method:
8b029fdf
MH
982
983 study;
984
985And in gdb do:
986
987 (gdb) break Perl_pp_study
988
9b22382a 989And then step until you hit what you're
73013070 990looking for. This works well in a loop
8b029fdf
MH
991if you want to only break at certain iterations:
992
993 for my $c (1..100) {
994 study if $c == 50;
995 }
996
997=head2 Using gdb to look at what the parser/lexer are doing
998
73013070 999If you want to see what perl is doing when parsing/lexing your code, you can
72b22e55 1000use C<BEGIN {}>:
8b029fdf
MH
1001
1002 print "Before\n";
1003 BEGIN { study; }
1004 print "After\n";
1005
1006And in gdb:
1007
1008 (gdb) break Perl_pp_study
1009
1010If you want to see what the parser/lexer is doing inside of C<if> blocks and
1011the like you need to be a little trickier:
1012
73013070 1013 if ($a && $b && do { BEGIN { study } 1 } && $c) { ... }
8b029fdf 1014
04c692a8
DR
1015=head1 SOURCE CODE STATIC ANALYSIS
1016
1017Various tools exist for analysing C source code B<statically>, as
9b22382a 1018opposed to B<dynamically>, that is, without executing the code. It is
04c692a8
DR
1019possible to detect resource leaks, undefined behaviour, type
1020mismatches, portability problems, code paths that would cause illegal
1021memory accesses, and other similar problems by just parsing the C code
1022and looking at the resulting graph, what does it tell about the
9b22382a 1023execution and data flows. As a matter of fact, this is exactly how C
04c692a8
DR
1024compilers know to give warnings about dubious code.
1025
1026=head2 lint, splint
1027
1028The good old C code quality inspector, C<lint>, is available in several
1029platforms, but please be aware that there are several different
1030implementations of it by different vendors, which means that the flags
1031are not identical across different platforms.
1032
1033There is a lint variant called C<splint> (Secure Programming Lint)
1034available from http://www.splint.org/ that should compile on any
1035Unix-like platform.
1036
1037There are C<lint> and <splint> targets in Makefile, but you may have to
1038diddle with the flags (see above).
1039
1040=head2 Coverity
1041
1042Coverity (http://www.coverity.com/) is a product similar to lint and as
1043a testbed for their product they periodically check several open source
1044projects, and they give out accounts to open source developers to the
1045defect databases.
1046
1047=head2 cpd (cut-and-paste detector)
1048
9b22382a 1049The cpd tool detects cut-and-paste coding. If one instance of the
04c692a8 1050cut-and-pasted code changes, all the other spots should probably be
9b22382a 1051changed, too. Therefore such code should probably be turned into a
04c692a8
DR
1052subroutine or a macro.
1053
1054cpd (http://pmd.sourceforge.net/cpd.html) is part of the pmd project
9b22382a 1055(http://pmd.sourceforge.net/). pmd was originally written for static
04c692a8
DR
1056analysis of Java code, but later the cpd part of it was extended to
1057parse also C and C++.
1058
1059Download the pmd-bin-X.Y.zip () from the SourceForge site, extract the
1060pmd-X.Y.jar from it, and then run that on source code thusly:
1061
0cbf2b31
FC
1062 java -cp pmd-X.Y.jar net.sourceforge.pmd.cpd.CPD \
1063 --minimum-tokens 100 --files /some/where/src --language c > cpd.txt
04c692a8
DR
1064
1065You may run into memory limits, in which case you should use the -Xmx
1066option:
1067
1068 java -Xmx512M ...
1069
1070=head2 gcc warnings
1071
1072Though much can be written about the inconsistency and coverage
1073problems of gcc warnings (like C<-Wall> not meaning "all the warnings",
1074or some common portability problems not being covered by C<-Wall>, or
1075C<-ansi> and C<-pedantic> both being a poorly defined collection of
1076warnings, and so forth), gcc is still a useful tool in keeping our
1077coding nose clean.
1078
1079The C<-Wall> is by default on.
1080
1081The C<-ansi> (and its sidekick, C<-pedantic>) would be nice to be on
1082always, but unfortunately they are not safe on all platforms, they can
1083for example cause fatal conflicts with the system headers (Solaris
9b22382a 1084being a prime example). If Configure C<-Dgccansipedantic> is used, the
04c692a8
DR
1085C<cflags> frontend selects C<-ansi -pedantic> for the platforms where
1086they are known to be safe.
1087
1088Starting from Perl 5.9.4 the following extra flags are added:
1089
1090=over 4
1091
1092=item *
1093
1094C<-Wendif-labels>
1095
1096=item *
1097
1098C<-Wextra>
1099
1100=item *
1101
1102C<-Wdeclaration-after-statement>
1103
1104=back
1105
1106The following flags would be nice to have but they would first need
1107their own Augean stablemaster:
1108
1109=over 4
1110
1111=item *
1112
1113C<-Wpointer-arith>
1114
1115=item *
1116
1117C<-Wshadow>
1118
1119=item *
1120
1121C<-Wstrict-prototypes>
1122
1123=back
1124
1125The C<-Wtraditional> is another example of the annoying tendency of gcc
1126to bundle a lot of warnings under one switch (it would be impossible to
1127deploy in practice because it would complain a lot) but it does contain
1128some warnings that would be beneficial to have available on their own,
1129such as the warning about string constants inside macros containing the
1130macro arguments: this behaved differently pre-ANSI than it does in
1131ANSI, and some C compilers are still in transition, AIX being an
1132example.
1133
1134=head2 Warnings of other C compilers
1135
1136Other C compilers (yes, there B<are> other C compilers than gcc) often
1137have their "strict ANSI" or "strict ANSI with some portability
1138extensions" modes on, like for example the Sun Workshop has its C<-Xa>
1139mode on (though implicitly), or the DEC (these days, HP...) has its
1140C<-std1> mode on.
1141
1142=head1 MEMORY DEBUGGERS
1143
d1fd4856
VP
1144B<NOTE 1>: Running under older memory debuggers such as Purify,
1145valgrind or Third Degree greatly slows down the execution: seconds
9b22382a 1146become minutes, minutes become hours. For example as of Perl 5.8.1, the
04c692a8 1147ext/Encode/t/Unicode.t takes extraordinarily long to complete under
9b22382a
FC
1148e.g. Purify, Third Degree, and valgrind. Under valgrind it takes more
1149than six hours, even on a snappy computer. The said test must be doing
1150something that is quite unfriendly for memory debuggers. If you don't
04c692a8 1151feel like waiting, that you can simply kill away the perl process.
d1fd4856
VP
1152Roughly valgrind slows down execution by factor 10, AddressSanitizer by
1153factor 2.
04c692a8
DR
1154
1155B<NOTE 2>: To minimize the number of memory leak false alarms (see
1156L</PERL_DESTRUCT_LEVEL> for more information), you have to set the
9b22382a 1157environment variable PERL_DESTRUCT_LEVEL to 2. For example, like this:
04c692a8
DR
1158
1159 env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib ...
1160
1161B<NOTE 3>: There are known memory leaks when there are compile-time
1162errors within eval or require, seeing C<S_doeval> in the call stack is
9b22382a 1163a good sign of these. Fixing these leaks is non-trivial, unfortunately,
04c692a8
DR
1164but they must be fixed eventually.
1165
1166B<NOTE 4>: L<DynaLoader> will not clean up after itself completely
1167unless Perl is built with the Configure option
1168C<-Accflags=-DDL_UNLOAD_ALL_AT_EXIT>.
1169
04c692a8
DR
1170=head2 valgrind
1171
d1fd4856 1172The valgrind tool can be used to find out both memory leaks and illegal
9b22382a 1173heap memory accesses. As of version 3.3.0, Valgrind only supports Linux
0263e49a 1174on x86, x86-64 and PowerPC and Darwin (OS X) on x86 and x86-64. The
d1fd4856 1175special "test.valgrind" target can be used to run the tests under
9b22382a 1176valgrind. Found errors and memory leaks are logged in files named
037ab3f1
MH
1177F<testfile.valgrind> and by default output is displayed inline.
1178
1179Example usage:
1180
1181 make test.valgrind
1182
1183Since valgrind adds significant overhead, tests will take much longer to
1184run. The valgrind tests support being run in parallel to help with this:
1185
1186 TEST_JOBS=9 make test.valgrind
1187
1188Note that the above two invocations will be very verbose as reachable
1189memory and leak-checking is enabled by default. If you want to just see
1190pure errors, try:
73013070 1191
037ab3f1
MH
1192 VG_OPTS='-q --leak-check=no --show-reachable=no' TEST_JOBS=9 \
1193 make test.valgrind
04c692a8
DR
1194
1195Valgrind also provides a cachegrind tool, invoked on perl as:
1196
1197 VG_OPTS=--tool=cachegrind make test.valgrind
1198
1199As system libraries (most notably glibc) are also triggering errors,
9b22382a 1200valgrind allows to suppress such errors using suppression files. The
04c692a8 1201default suppression file that comes with valgrind already catches a lot
9b22382a 1202of them. Some additional suppressions are defined in F<t/perl.supp>.
04c692a8
DR
1203
1204To get valgrind and for more information see
1205
0061d4fa 1206 http://valgrind.org/
04c692a8 1207
81c3bbe7
RU
1208=head2 AddressSanitizer
1209
4dd56148 1210AddressSanitizer is a clang and gcc extension, included in clang since
9b22382a 1211v3.1 and gcc since v4.8. It checks illegal heap pointers, global
4dd56148
NC
1212pointers, stack pointers and use after free errors, and is fast enough
1213that you can easily compile your debugging or optimized perl with it.
9b22382a 1214It does not check memory leaks though. AddressSanitizer is available
4dd56148 1215for Linux, Mac OS X and soon on Windows.
81c3bbe7 1216
8a64fbaa
VP
1217To build perl with AddressSanitizer, your Configure invocation should
1218look like:
81c3bbe7 1219
e8596d90
VP
1220 sh Configure -des -Dcc=clang \
1221 -Accflags=-faddress-sanitizer -Aldflags=-faddress-sanitizer \
1222 -Alddlflags=-shared\ -faddress-sanitizer
81c3bbe7
RU
1223
1224where these arguments mean:
1225
1226=over 4
1227
1228=item * -Dcc=clang
1229
8a64fbaa
VP
1230This should be replaced by the full path to your clang executable if it
1231is not in your path.
81c3bbe7
RU
1232
1233=item * -Accflags=-faddress-sanitizer
1234
8a64fbaa 1235Compile perl and extensions sources with AddressSanitizer.
81c3bbe7
RU
1236
1237=item * -Aldflags=-faddress-sanitizer
1238
8a64fbaa 1239Link the perl executable with AddressSanitizer.
81c3bbe7 1240
e8596d90 1241=item * -Alddlflags=-shared\ -faddress-sanitizer
81c3bbe7 1242
9b22382a 1243Link dynamic extensions with AddressSanitizer. You must manually
e8596d90
VP
1244specify C<-shared> because using C<-Alddlflags=-shared> will prevent
1245Configure from setting a default value for C<lddlflags>, which usually
5dfc6e97 1246contains C<-shared> (at least on Linux).
81c3bbe7
RU
1247
1248=back
1249
8a64fbaa
VP
1250See also
1251L<http://code.google.com/p/address-sanitizer/wiki/AddressSanitizer>.
81c3bbe7
RU
1252
1253
04c692a8
DR
1254=head1 PROFILING
1255
1256Depending on your platform there are various ways of profiling Perl.
1257
1258There are two commonly used techniques of profiling executables:
1259I<statistical time-sampling> and I<basic-block counting>.
1260
1261The first method takes periodically samples of the CPU program counter,
1262and since the program counter can be correlated with the code generated
1263for functions, we get a statistical view of in which functions the
9b22382a 1264program is spending its time. The caveats are that very small/fast
04c692a8
DR
1265functions have lower probability of showing up in the profile, and that
1266periodically interrupting the program (this is usually done rather
1267frequently, in the scale of milliseconds) imposes an additional
9b22382a 1268overhead that may skew the results. The first problem can be alleviated
04c692a8
DR
1269by running the code for longer (in general this is a good idea for
1270profiling), the second problem is usually kept in guard by the
1271profiling tools themselves.
1272
1273The second method divides up the generated code into I<basic blocks>.
1274Basic blocks are sections of code that are entered only in the
9b22382a
FC
1275beginning and exited only at the end. For example, a conditional jump
1276starts a basic block. Basic block profiling usually works by
04c692a8 1277I<instrumenting> the code by adding I<enter basic block #nnnn>
9b22382a
FC
1278book-keeping code to the generated code. During the execution of the
1279code the basic block counters are then updated appropriately. The
04c692a8
DR
1280caveat is that the added extra code can skew the results: again, the
1281profiling tools usually try to factor their own effects out of the
1282results.
1283
1284=head2 Gprof Profiling
1285
e2aed43d 1286I<gprof> is a profiling tool available in many Unix platforms which
9b22382a
FC
1287uses I<statistical time-sampling>. You can build a profiled version of
1288F<perl> by compiling using gcc with the flag C<-pg>. Either edit
1289F<config.sh> or re-run F<Configure>. Running the profiled version of
e2aed43d
NC
1290Perl will create an output file called F<gmon.out> which contains the
1291profiling data collected during the execution.
04c692a8 1292
e2aed43d
NC
1293quick hint:
1294
1295 $ sh Configure -des -Dusedevel -Accflags='-pg' \
1296 -Aldflags='-pg' -Alddlflags='-pg -shared' \
1297 && make perl
1298 $ ./perl ... # creates gmon.out in current directory
1299 $ gprof ./perl > out
1300 $ less out
1301
1302(you probably need to add C<-shared> to the <-Alddlflags> line until RT
1303#118199 is resolved)
04c692a8 1304
e2aed43d
NC
1305The F<gprof> tool can then display the collected data in various ways.
1306Usually F<gprof> understands the following options:
04c692a8
DR
1307
1308=over 4
1309
1310=item * -a
1311
1312Suppress statically defined functions from the profile.
1313
1314=item * -b
1315
1316Suppress the verbose descriptions in the profile.
1317
1318=item * -e routine
1319
1320Exclude the given routine and its descendants from the profile.
1321
1322=item * -f routine
1323
1324Display only the given routine and its descendants in the profile.
1325
1326=item * -s
1327
1328Generate a summary file called F<gmon.sum> which then may be given to
1329subsequent gprof runs to accumulate data over several runs.
1330
1331=item * -z
1332
1333Display routines that have zero usage.
1334
1335=back
1336
1337For more detailed explanation of the available commands and output
e2aed43d 1338formats, see your own local documentation of F<gprof>.
04c692a8 1339
e2aed43d 1340=head2 GCC gcov Profiling
04c692a8 1341
e2aed43d
NC
1342I<basic block profiling> is officially available in gcc 3.0 and later.
1343You can build a profiled version of F<perl> by compiling using gcc with
9b22382a 1344the flags C<-fprofile-arcs -ftest-coverage>. Either edit F<config.sh>
e2aed43d 1345or re-run F<Configure>.
04c692a8 1346
e2aed43d 1347quick hint:
04c692a8 1348
e2aed43d
NC
1349 $ sh Configure -des -Dusedevel -Doptimize='-g' \
1350 -Accflags='-fprofile-arcs -ftest-coverage' \
1351 -Aldflags='-fprofile-arcs -ftest-coverage' \
1352 -Alddlflags='-fprofile-arcs -ftest-coverage -shared' \
1353 && make perl
1354 $ rm -f regexec.c.gcov regexec.gcda
1355 $ ./perl ...
1356 $ gcov regexec.c
1357 $ less regexec.c.gcov
04c692a8 1358
e2aed43d
NC
1359(you probably need to add C<-shared> to the <-Alddlflags> line until RT
1360#118199 is resolved)
04c692a8
DR
1361
1362Running the profiled version of Perl will cause profile output to be
9b22382a 1363generated. For each source file an accompanying F<.gcda> file will be
04c692a8
DR
1364created.
1365
e2aed43d 1366To display the results you use the I<gcov> utility (which should be
9b22382a 1367installed if you have gcc 3.0 or newer installed). F<gcov> is run on
04c692a8
DR
1368source code files, like this
1369
1370 gcov sv.c
1371
9b22382a 1372which will cause F<sv.c.gcov> to be created. The F<.gcov> files contain
04c692a8 1373the source code annotated with relative frequencies of execution
9b22382a 1374indicated by "#" markers. If you want to generate F<.gcov> files for
6f134219
NC
1375all profiled object files, you can run something like this:
1376
1377 for file in `find . -name \*.gcno`
1378 do sh -c "cd `dirname $file` && gcov `basename $file .gcno`"
1379 done
04c692a8
DR
1380
1381Useful options of F<gcov> include C<-b> which will summarise the basic
1382block, branch, and function call coverage, and C<-c> which instead of
9b22382a 1383relative frequencies will use the actual counts. For more information
04c692a8 1384on the use of F<gcov> and basic block profiling with gcc, see the
9b22382a 1385latest GNU CC manual. As of gcc 4.8, this is at
e2aed43d 1386L<http://gcc.gnu.org/onlinedocs/gcc/Gcov-Intro.html#Gcov-Intro>
04c692a8
DR
1387
1388=head1 MISCELLANEOUS TRICKS
1389
1390=head2 PERL_DESTRUCT_LEVEL
1391
1392If you want to run any of the tests yourself manually using e.g.
4dd56148
NC
1393valgrind, please note that by default perl B<does not> explicitly
1394cleanup all the memory it has allocated (such as global memory arenas)
1395but instead lets the exit() of the whole program "take care" of such
1396allocations, also known as "global destruction of objects".
04c692a8
DR
1397
1398There is a way to tell perl to do complete cleanup: set the environment
9b22382a 1399variable PERL_DESTRUCT_LEVEL to a non-zero value. The t/TEST wrapper
04c692a8 1400does set this to 2, and this is what you need to do too, if you don't
f01ecde8 1401want to see the "global leaks": For example, for running under valgrind
04c692a8 1402
a63ef199 1403 env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib t/foo/bar.t
04c692a8
DR
1404
1405(Note: the mod_perl apache module uses also this environment variable
9b22382a
FC
1406for its own purposes and extended its semantics. Refer to the mod_perl
1407documentation for more information. Also, spawned threads do the
04c692a8
DR
1408equivalent of setting this variable to the value 1.)
1409
1410If, at the end of a run you get the message I<N scalars leaked>, you
1411can recompile with C<-DDEBUG_LEAKING_SCALARS>, which will cause the
1412addresses of all those leaked SVs to be dumped along with details as to
9b22382a
FC
1413where each SV was originally allocated. This information is also
1414displayed by Devel::Peek. Note that the extra details recorded with
04c692a8 1415each SV increases memory usage, so it shouldn't be used in production
9b22382a 1416environments. It also converts C<new_SV()> from a macro into a real
04c692a8
DR
1417function, so you can use your favourite debugger to discover where
1418those pesky SVs were allocated.
1419
1420If you see that you're leaking memory at runtime, but neither valgrind
1421nor C<-DDEBUG_LEAKING_SCALARS> will find anything, you're probably
1422leaking SVs that are still reachable and will be properly cleaned up
9b22382a
FC
1423during destruction of the interpreter. In such cases, using the C<-Dm>
1424switch can point you to the source of the leak. If the executable was
04c692a8 1425built with C<-DDEBUG_LEAKING_SCALARS>, C<-Dm> will output SV
9b22382a 1426allocations in addition to memory allocations. Each SV allocation has a
04c692a8 1427distinct serial number that will be written on creation and destruction
9b22382a 1428of the SV. So if you're executing the leaking code in a loop, you need
04c692a8 1429to look for SVs that are created, but never destroyed between each
9b22382a 1430cycle. If such an SV is found, set a conditional breakpoint within
04c692a8 1431C<new_SV()> and make it break only when C<PL_sv_serial> is equal to the
9b22382a 1432serial number of the leaking SV. Then you will catch the interpreter in
04c692a8
DR
1433exactly the state where the leaking SV is allocated, which is
1434sufficient in many cases to find the source of the leak.
1435
1436As C<-Dm> is using the PerlIO layer for output, it will by itself
9b22382a 1437allocate quite a bunch of SVs, which are hidden to avoid recursion. You
04c692a8
DR
1438can bypass the PerlIO layer if you use the SV logging provided by
1439C<-DPERL_MEM_LOG> instead.
1440
1441=head2 PERL_MEM_LOG
1442
1443If compiled with C<-DPERL_MEM_LOG>, both memory and SV allocations go
1444through logging functions, which is handy for breakpoint setting.
1445
1446Unless C<-DPERL_MEM_LOG_NOIMPL> is also compiled, the logging functions
1447read $ENV{PERL_MEM_LOG} to determine whether to log the event, and if
1448so how:
1449
a63ef199
SF
1450 $ENV{PERL_MEM_LOG} =~ /m/ Log all memory ops
1451 $ENV{PERL_MEM_LOG} =~ /s/ Log all SV ops
1452 $ENV{PERL_MEM_LOG} =~ /t/ include timestamp in Log
1453 $ENV{PERL_MEM_LOG} =~ /^(\d+)/ write to FD given (default is 2)
04c692a8
DR
1454
1455Memory logging is somewhat similar to C<-Dm> but is independent of
1456C<-DDEBUGGING>, and at a higher level; all uses of Newx(), Renew(), and
1457Safefree() are logged with the caller's source code file and line
9b22382a
FC
1458number (and C function name, if supported by the C compiler). In
1459contrast, C<-Dm> is directly at the point of C<malloc()>. SV logging is
04c692a8
DR
1460similar.
1461
1462Since the logging doesn't use PerlIO, all SV allocations are logged and
9b22382a 1463no extra SV allocations are introduced by enabling the logging. If
04c692a8
DR
1464compiled with C<-DDEBUG_LEAKING_SCALARS>, the serial number for each SV
1465allocation is also logged.
1466
1467=head2 DDD over gdb
1468
1469Those debugging perl with the DDD frontend over gdb may find the
1470following useful:
1471
1472You can extend the data conversion shortcuts menu, so for example you
1473can display an SV's IV value with one click, without doing any typing.
1474To do that simply edit ~/.ddd/init file and add after:
1475
1476 ! Display shortcuts.
1477 Ddd*gdbDisplayShortcuts: \
1478 /t () // Convert to Bin\n\
1479 /d () // Convert to Dec\n\
1480 /x () // Convert to Hex\n\
1481 /o () // Convert to Oct(\n\
1482
1483the following two lines:
1484
1485 ((XPV*) (())->sv_any )->xpv_pv // 2pvx\n\
1486 ((XPVIV*) (())->sv_any )->xiv_iv // 2ivx
1487
1488so now you can do ivx and pvx lookups or you can plug there the sv_peek
1489"conversion":
1490
1491 Perl_sv_peek(my_perl, (SV*)()) // sv_peek
1492
9b22382a 1493(The my_perl is for threaded builds.) Just remember that every line,
04c692a8
DR
1494but the last one, should end with \n\
1495
1496Alternatively edit the init file interactively via: 3rd mouse button ->
1497New Display -> Edit Menu
1498
1499Note: you can define up to 20 conversion shortcuts in the gdb section.
1500
470dd224
JH
1501=head2 C backtrace
1502
0762e42f
JH
1503On some platforms Perl supports retrieving the C level backtrace
1504(similar to what symbolic debuggers like gdb do).
470dd224
JH
1505
1506The backtrace returns the stack trace of the C call frames,
1507with the symbol names (function names), the object names (like "perl"),
1508and if it can, also the source code locations (file:line).
1509
0762e42f
JH
1510The supported platforms are Linux, and OS X (some *BSD might
1511work at least partly, but they have not yet been tested).
1512
1513This feature hasn't been tested with multiple threads, but it will
1514only show the backtrace of the thread doing the backtracing.
470dd224
JH
1515
1516The feature needs to be enabled with C<Configure -Dusecbacktrace>.
1517
0762e42f
JH
1518The C<-Dusecbacktrace> also enables keeping the debug information when
1519compiling/linking (often: C<-g>). Many compilers/linkers do support
1520having both optimization and keeping the debug information. The debug
1521information is needed for the symbol names and the source locations.
1522
1523Static functions might not be visible for the backtrace.
470dd224
JH
1524
1525Source code locations, even if available, can often be missing or
0762e42f
JH
1526misleading if the compiler has e.g. inlined code. Optimizer can
1527make matching the source code and the object code quite challenging.
470dd224
JH
1528
1529=over 4
1530
1531=item Linux
1532
59b3baca 1533You B<must> have the BFD (-lbfd) library installed, otherwise C<perl> will
0762e42f 1534fail to link. The BFD is usually distributed as part of the GNU binutils.
470dd224
JH
1535
1536Summary: C<Configure ... -Dusecbacktrace>
1537and you need C<-lbfd>.
1538
1539=item OS X
1540
0762e42f
JH
1541The source code locations are supported B<only> if you have
1542the Developer Tools installed. (BFD is B<not> needed.)
470dd224
JH
1543
1544Summary: C<Configure ... -Dusecbacktrace>
1545and installing the Developer Tools would be good.
1546
1547=back
1548
1549Optionally, for trying out the feature, you may want to enable
0762e42f
JH
1550automatic dumping of the backtrace just before a warning or croak (die)
1551message is emitted, by adding C<-Accflags=-DUSE_C_BACKTRACE_ON_ERROR>
1552for Configure.
470dd224
JH
1553
1554Unless the above additional feature is enabled, nothing about the
1555backtrace functionality is visible, except for the Perl/XS level.
1556
1557Furthermore, even if you have enabled this feature to be compiled,
1558you need to enable it in runtime with an environment variable:
0762e42f
JH
1559C<PERL_C_BACKTRACE_ON_ERROR=10>. It must be an integer higher
1560than zero, telling the desired frame count.
470dd224
JH
1561
1562Retrieving the backtrace from Perl level (using for example an XS
1563extension) would be much less exciting than one would hope: normally
1564you would see C<runops>, C<entersub>, and not much else. This API is
1565intended to be called B<from within> the Perl implementation, not from
1566Perl level execution.
1567
0762e42f 1568The C API for the backtrace is as follows:
470dd224
JH
1569
1570=over 4
1571
1572=item get_c_backtrace
1573
1574=item free_c_backtrace
1575
1576=item get_c_backtrace_dump
1577
1578=item dump_c_backtrace
1579
1580=back
1581
04c692a8
DR
1582=head2 Poison
1583
1584If you see in a debugger a memory area mysteriously full of 0xABABABAB
1585or 0xEFEFEFEF, you may be seeing the effect of the Poison() macros, see
1586L<perlclib>.
1587
1588=head2 Read-only optrees
1589
9b22382a 1590Under ithreads the optree is read only. If you want to enforce this, to
04c692a8 1591check for write accesses from buggy code, compile with
91fc0422
FC
1592C<-Accflags=-DPERL_DEBUG_READONLY_OPS>
1593to enable code that allocates op memory
4dd56148
NC
1594via C<mmap>, and sets it read-only when it is attached to a subroutine.
1595Any write access to an op results in a C<SIGBUS> and abort.
04c692a8
DR
1596
1597This code is intended for development only, and may not be portable
9b22382a
FC
1598even to all Unix variants. Also, it is an 80% solution, in that it
1599isn't able to make all ops read only. Specifically it does not apply to
4dd56148 1600op slabs belonging to C<BEGIN> blocks.
04c692a8 1601
4dd56148
NC
1602However, as an 80% solution it is still effective, as it has caught
1603bugs in the past.
04c692a8 1604
f789f6a4
FC
1605=head2 When is a bool not a bool?
1606
1607On pre-C99 compilers, C<bool> is defined as equivalent to C<char>.
1608Consequently assignment of any larger type to a C<bool> is unsafe and may
1609be truncated. The C<cBOOL> macro exists to cast it correctly.
1610
1611On those platforms and compilers where C<bool> really is a boolean (C++,
1612C99), it is easy to forget the cast. You can force C<bool> to be a C<char>
1613by compiling with C<-Accflags=-DPERL_BOOL_AS_CHAR>. You may also wish to
50e4f4d4
CB
1614run C<Configure> with something like
1615
cbc13c3d 1616 -Accflags='-Wconversion -Wno-sign-conversion -Wno-shorten-64-to-32'
50e4f4d4
CB
1617
1618or your compiler's equivalent to make it easier to spot any unsafe truncations
1619that show up.
f789f6a4 1620
04c692a8
DR
1621=head2 The .i Targets
1622
1623You can expand the macros in a F<foo.c> file by saying
1624
1625 make foo.i
1626
d1fd4856
VP
1627which will expand the macros using cpp. Don't be scared by the
1628results.
04c692a8
DR
1629
1630=head1 AUTHOR
1631
1632This document was originally written by Nathan Torkington, and is
1633maintained by the perl5-porters mailing list.