This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Eliminate opASSIGN macro usage from core
[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
adf7d503 23Perl source plays by ANSI C89 rules: no C99 (or C++) extensions.
04c692a8
DR
24You don't care about some particular platform having broken Perl? I
25hear there is still a strong demand for J2EE programmers.
26
27=head2 Perl environment problems
28
29=over 4
30
31=item *
32
33Not compiling with threading
34
35Compiling with threading (-Duseithreads) completely rewrites the
9b22382a 36function prototypes of Perl. You better try your changes with that.
04c692a8
DR
37Related to this is the difference between "Perl_-less" and "Perl_-ly"
38APIs, for example:
39
40 Perl_sv_setiv(aTHX_ ...);
41 sv_setiv(...);
42
43The first one explicitly passes in the context, which is needed for
9b22382a
FC
44e.g. threaded builds. The second one does that implicitly; do not get
45them mixed. If you are not passing in a aTHX_, you will need to do a
04c692a8
DR
46dTHX (or a dVAR) as the first thing in the function.
47
48See L<perlguts/"How multiple interpreters and concurrency are
49supported"> for further discussion about context.
50
51=item *
52
53Not compiling with -DDEBUGGING
54
55The DEBUGGING define exposes more code to the compiler, therefore more
9b22382a 56ways for things to go wrong. You should try it.
04c692a8
DR
57
58=item *
59
60Introducing (non-read-only) globals
61
62Do not introduce any modifiable globals, truly global or file static.
63They are bad form and complicate multithreading and other forms of
9b22382a 64concurrency. The right way is to introduce them as new interpreter
04c692a8
DR
65variables, see F<intrpvar.h> (at the very end for binary
66compatibility).
67
68Introducing read-only (const) globals is okay, as long as you verify
69with e.g. C<nm libperl.a|egrep -v ' [TURtr] '> (if your C<nm> has
9b22382a 70BSD-style output) that the data you added really is read-only. (If it
04c692a8
DR
71is, it shouldn't show up in the output of that command.)
72
73If you want to have static strings, make them constant:
74
75 static const char etc[] = "...";
76
77If you want to have arrays of constant strings, note carefully the
78right combination of C<const>s:
79
80 static const char * const yippee[] =
a63ef199 81 {"hi", "ho", "silver"};
04c692a8
DR
82
83There is a way to completely hide any modifiable globals (they are all
84moved to heap), the compilation setting
9b22382a 85C<-DPERL_GLOBAL_STRUCT_PRIVATE>. It is not normally used, but can be
04c692a8
DR
86used for testing, read more about it in L<perlguts/"Background and
87PERL_IMPLICIT_CONTEXT">.
88
89=item *
90
91Not exporting your new function
92
93Some platforms (Win32, AIX, VMS, OS/2, to name a few) require any
94function that is part of the public API (the shared Perl library) to be
9b22382a 95explicitly marked as exported. See the discussion about F<embed.pl> in
04c692a8
DR
96L<perlguts>.
97
98=item *
99
100Exporting your new function
101
102The new shiny result of either genuine new functionality or your
9b22382a 103arduous refactoring is now ready and correctly exported. So what could
04c692a8
DR
104possibly go wrong?
105
106Maybe simply that your function did not need to be exported in the
9b22382a 107first place. Perl has a long and not so glorious history of exporting
04c692a8
DR
108functions that it should not have.
109
110If the function is used only inside one source code file, make it
9b22382a 111static. See the discussion about F<embed.pl> in L<perlguts>.
04c692a8
DR
112
113If the function is used across several files, but intended only for
114Perl's internal use (and this should be the common case), do not export
9b22382a 115it to the public API. See the discussion about F<embed.pl> in
04c692a8
DR
116L<perlguts>.
117
118=back
119
120=head2 Portability problems
121
122The following are common causes of compilation and/or execution
9b22382a
FC
123failures, not common to Perl as such. The C FAQ is good bedtime
124reading. Please test your changes with as many C compilers and
04c692a8
DR
125platforms as possible; we will, anyway, and it's nice to save oneself
126from public embarrassment.
127
128If using gcc, you can add the C<-std=c89> option which will hopefully
9b22382a 129catch most of these unportabilities. (However it might also catch
04c692a8
DR
130incompatibilities in your system's header files.)
131
132Use the Configure C<-Dgccansipedantic> flag to enable the gcc C<-ansi
133-pedantic> flags which enforce stricter ANSI rules.
134
135If using the C<gcc -Wall> note that not all the possible warnings (like
404b28f3 136C<-Wuninitialized>) are given unless you also compile with C<-O>.
04c692a8
DR
137
138Note that if using gcc, starting from Perl 5.9.5 the Perl core source
139code files (the ones at the top level of the source code distribution,
140but not e.g. the extensions under ext/) are automatically compiled with
141as many as possible of the C<-std=c89>, C<-ansi>, C<-pedantic>, and a
142selection of C<-W> flags (see cflags.SH).
143
144Also study L<perlport> carefully to avoid any bad assumptions about the
eb9df707 145operating system, filesystems, character set, and so forth.
04c692a8
DR
146
147You may once in a while try a "make microperl" to see whether we can
9b22382a 148still compile Perl with just the bare minimum of interfaces. (See
04c692a8
DR
149README.micro.)
150
151Do not assume an operating system indicates a certain compiler.
152
153=over 4
154
155=item *
156
157Casting pointers to integers or casting integers to pointers
158
159 void castaway(U8* p)
160 {
161 IV i = p;
162
163or
164
165 void castaway(U8* p)
166 {
167 IV i = (IV)p;
168
9b22382a
FC
169Both are bad, and broken, and unportable. Use the PTR2IV() macro that
170does it right. (Likewise, there are PTR2UV(), PTR2NV(), INT2PTR(), and
04c692a8
DR
171NUM2PTR().)
172
173=item *
174
28ffebaf 175Casting between function pointers and data pointers
04c692a8
DR
176
177Technically speaking casting between function pointers and data
178pointers is unportable and undefined, but practically speaking it seems
179to work, but you should use the FPTR2DPTR() and DPTR2FPTR() macros.
180Sometimes you can also play games with unions.
181
182=item *
183
184Assuming sizeof(int) == sizeof(long)
185
186There are platforms where longs are 64 bits, and platforms where ints
187are 64 bits, and while we are out to shock you, even platforms where
9b22382a 188shorts are 64 bits. This is all legal according to the C standard. (In
04c692a8
DR
189other words, "long long" is not a portable way to specify 64 bits, and
190"long long" is not even guaranteed to be any wider than "long".)
191
192Instead, use the definitions IV, UV, IVSIZE, I32SIZE, and so forth.
193Avoid things like I32 because they are B<not> guaranteed to be
194I<exactly> 32 bits, they are I<at least> 32 bits, nor are they
9b22382a 195guaranteed to be B<int> or B<long>. If you really explicitly need
04c692a8
DR
19664-bit variables, use I64 and U64, but only if guarded by HAS_QUAD.
197
198=item *
199
200Assuming one can dereference any type of pointer for any type of data
201
202 char *p = ...;
56bb4b7b 203 long pony = *(long *)p; /* BAD */
04c692a8
DR
204
205Many platforms, quite rightly so, will give you a core dump instead of
768312ab 206a pony if the p happens not to be correctly aligned.
04c692a8
DR
207
208=item *
209
210Lvalue casts
211
212 (int)*p = ...; /* BAD */
213
9b22382a 214Simply not portable. Get your lvalue to be of the right type, or maybe
04c692a8
DR
215use temporary variables, or dirty tricks with unions.
216
217=item *
218
219Assume B<anything> about structs (especially the ones you don't
220control, like the ones coming from the system headers)
221
222=over 8
223
224=item *
225
226That a certain field exists in a struct
227
228=item *
229
230That no other fields exist besides the ones you know of
231
232=item *
233
234That a field is of certain signedness, sizeof, or type
235
236=item *
237
238That the fields are in a certain order
239
240=over 8
241
242=item *
243
244While C guarantees the ordering specified in the struct definition,
245between different platforms the definitions might differ
246
247=back
248
249=item *
250
251That the sizeof(struct) or the alignments are the same everywhere
252
253=over 8
254
255=item *
256
257There might be padding bytes between the fields to align the fields -
258the bytes can be anything
259
260=item *
261
262Structs are required to be aligned to the maximum alignment required by
263the fields - which for native types is for usually equivalent to
264sizeof() of the field
265
266=back
267
268=back
269
270=item *
271
272Assuming the character set is ASCIIish
273
9b22382a 274Perl can compile and run under EBCDIC platforms. See L<perlebcdic>.
04c692a8
DR
275This is transparent for the most part, but because the character sets
276differ, you shouldn't use numeric (decimal, octal, nor hex) constants
eb9df707
KW
277to refer to characters. You can safely say C<'A'>, but not C<0x41>.
278You can safely say C<'\n'>, but not C<\012>. However, you can use
279macros defined in F<utf8.h> to specify any code point portably.
280C<LATIN1_TO_NATIVE(0xDF)> is going to be the code point that means
281LATIN SMALL LETTER SHARP S on whatever platform you are running on (on
282ASCII platforms it compiles without adding any extra code, so there is
283zero performance hit on those). The acceptable inputs to
284C<LATIN1_TO_NATIVE> are from C<0x00> through C<0xFF>. If your input
285isn't guaranteed to be in that range, use C<UNICODE_TO_NATIVE> instead.
286C<NATIVE_TO_LATIN1> and C<NATIVE_TO_UNICODE> translate the opposite
287direction.
288
289If you need the string representation of a character that doesn't have a
290mnemonic name in C, you should add it to the list in
c22aa07d 291F<regen/unicode_constants.pl>, and have Perl create C<#define>'s for you,
eb6d698b 292based on the current platform.
04c692a8 293
eb9df707
KW
294Note that the C<isI<FOO>> and C<toI<FOO>> macros in F<handy.h> work
295properly on native code points and strings.
296
04c692a8 297Also, the range 'A' - 'Z' in ASCII is an unbroken sequence of 26 upper
9b22382a
FC
298case alphabetic characters. That is not true in EBCDIC. Nor for 'a' to
299'z'. But '0' - '9' is an unbroken range in both systems. Don't assume
c22aa07d 300anything about other ranges. (Note that special handling of ranges in
f4240379 301regular expression patterns and transliterations makes it appear to Perl
c22aa07d 302code that the aforementioned ranges are all unbroken.)
04c692a8
DR
303
304Many of the comments in the existing code ignore the possibility of
9b22382a 305EBCDIC, and may be wrong therefore, even if the code works. This is
04c692a8
DR
306actually a tribute to the successful transparent insertion of being
307able to handle EBCDIC without having to change pre-existing code.
308
309UTF-8 and UTF-EBCDIC are two different encodings used to represent
9b22382a 310Unicode code points as sequences of bytes. Macros with the same names
eb9df707 311(but different definitions) in F<utf8.h> and F<utfebcdic.h> are used to
04c692a8
DR
312allow the calling code to think that there is only one such encoding.
313This is almost always referred to as C<utf8>, but it means the EBCDIC
9b22382a 314version as well. Again, comments in the code may well be wrong even if
eb9df707 315the code itself is right. For example, the concept of UTF-8 C<invariant
9b22382a
FC
316characters> differs between ASCII and EBCDIC. On ASCII platforms, only
317characters that do not have the high-order bit set (i.e. whose ordinals
04c692a8
DR
318are strict ASCII, 0 - 127) are invariant, and the documentation and
319comments in the code may assume that, often referring to something
9b22382a 320like, say, C<hibit>. The situation differs and is not so simple on
04c692a8
DR
321EBCDIC machines, but as long as the code itself uses the
322C<NATIVE_IS_INVARIANT()> macro appropriately, it works, even if the
323comments are wrong.
324
257844b9
KW
325As noted in L<perlhack/TESTING>, when writing test scripts, the file
326F<t/charset_tools.pl> contains some helpful functions for writing tests
327valid on both ASCII and EBCDIC platforms. Sometimes, though, a test
328can't use a function and it's inconvenient to have different test
329versions depending on the platform. There are 20 code points that are
330the same in all 4 character sets currently recognized by Perl (the 3
331EBCDIC code pages plus ISO 8859-1 (ASCII/Latin1)). These can be used in
332such tests, though there is a small possibility that Perl will become
333available in yet another character set, breaking your test. All but one
334of these code points are C0 control characters. The most significant
335controls that are the same are C<\0>, C<\r>, and C<\N{VT}> (also
336specifiable as C<\cK>, C<\x0B>, C<\N{U+0B}>, or C<\013>). The single
337non-control is U+00B6 PILCROW SIGN. The controls that are the same have
338the same bit pattern in all 4 character sets, regardless of the UTF8ness
339of the string containing them. The bit pattern for U+B6 is the same in
340all 4 for non-UTF8 strings, but differs in each when its containing
341string is UTF-8 encoded. The only other code points that have some sort
342of sameness across all 4 character sets are the pair 0xDC and 0xFC.
343Together these represent upper- and lowercase LATIN LETTER U WITH
344DIAERESIS, but which is upper and which is lower may be reversed: 0xDC
345is the capital in Latin1 and 0xFC is the small letter, while 0xFC is the
346capital in EBCDIC and 0xDC is the small one. This factoid may be
347exploited in writing case insensitive tests that are the same across all
3484 character sets.
349
04c692a8
DR
350=item *
351
352Assuming the character set is just ASCII
353
9b22382a 354ASCII is a 7 bit encoding, but bytes have 8 bits in them. The 128 extra
04c692a8
DR
355characters have different meanings depending on the locale. Absent a
356locale, currently these extra characters are generally considered to be
eb9df707
KW
357unassigned, and this has presented some problems. This has being
358changed starting in 5.12 so that these characters can be considered to
359be Latin-1 (ISO-8859-1).
04c692a8
DR
360
361=item *
362
363Mixing #define and #ifdef
364
365 #define BURGLE(x) ... \
366 #ifdef BURGLE_OLD_STYLE /* BAD */
367 ... do it the old way ... \
368 #else
369 ... do it the new way ... \
370 #endif
371
9b22382a 372You cannot portably "stack" cpp directives. For example in the above
04c692a8
DR
373you need two separate BURGLE() #defines, one for each #ifdef branch.
374
375=item *
376
377Adding non-comment stuff after #endif or #else
378
379 #ifdef SNOSH
380 ...
381 #else !SNOSH /* BAD */
382 ...
383 #endif SNOSH /* BAD */
384
385The #endif and #else cannot portably have anything non-comment after
9b22382a 386them. If you want to document what is going (which is a good idea
04c692a8
DR
387especially if the branches are long), use (C) comments:
388
389 #ifdef SNOSH
390 ...
391 #else /* !SNOSH */
392 ...
393 #endif /* SNOSH */
394
395The gcc option C<-Wendif-labels> warns about the bad variant (by
396default on starting from Perl 5.9.4).
397
398=item *
399
400Having a comma after the last element of an enum list
401
402 enum color {
403 CERULEAN,
404 CHARTREUSE,
405 CINNABAR, /* BAD */
406 };
407
9b22382a 408is not portable. Leave out the last comma.
04c692a8
DR
409
410Also note that whether enums are implicitly morphable to ints varies
411between compilers, you might need to (int).
412
413=item *
414
415Using //-comments
416
417 // This function bamfoodles the zorklator. /* BAD */
418
9b22382a 419That is C99 or C++. Perl is C89. Using the //-comments is silently
04c692a8
DR
420allowed by many C compilers but cranking up the ANSI C89 strictness
421(which we like to do) causes the compilation to fail.
422
423=item *
424
425Mixing declarations and code
426
427 void zorklator()
428 {
429 int n = 3;
430 set_zorkmids(n); /* BAD */
431 int q = 4;
432
9b22382a 433That is C99 or C++. Some C compilers allow that, but you shouldn't.
04c692a8
DR
434
435The gcc option C<-Wdeclaration-after-statements> scans for such
436problems (by default on starting from Perl 5.9.4).
437
438=item *
439
440Introducing variables inside for()
441
442 for(int i = ...; ...; ...) { /* BAD */
443
9b22382a 444That is C99 or C++. While it would indeed be awfully nice to have that
04c692a8
DR
445also in C89, to limit the scope of the loop variable, alas, we cannot.
446
447=item *
448
449Mixing signed char pointers with unsigned char pointers
450
451 int foo(char *s) { ... }
452 ...
453 unsigned char *t = ...; /* Or U8* t = ... */
454 foo(t); /* BAD */
455
456While this is legal practice, it is certainly dubious, and downright
457fatal in at least one platform: for example VMS cc considers this a
9b22382a 458fatal error. One cause for people often making this mistake is that a
04c692a8
DR
459"naked char" and therefore dereferencing a "naked char pointer" have an
460undefined signedness: it depends on the compiler and the flags of the
461compiler and the underlying platform whether the result is signed or
9b22382a 462unsigned. For this very same reason using a 'char' as an array index is
04c692a8
DR
463bad.
464
465=item *
466
467Macros that have string constants and their arguments as substrings of
468the string constants
469
470 #define FOO(n) printf("number = %d\n", n) /* BAD */
471 FOO(10);
472
473Pre-ANSI semantics for that was equivalent to
474
475 printf("10umber = %d\10");
476
9b22382a 477which is probably not what you were expecting. Unfortunately at least
04c692a8
DR
478one reasonably common and modern C compiler does "real backward
479compatibility" here, in AIX that is what still happens even though the
480rest of the AIX compiler is very happily C89.
481
482=item *
483
484Using printf formats for non-basic C types
485
486 IV i = ...;
487 printf("i = %d\n", i); /* BAD */
488
489While this might by accident work in some platform (where IV happens to
9b22382a 490be an C<int>), in general it cannot. IV might be something larger. Even
04c692a8
DR
491worse the situation is with more specific types (defined by Perl's
492configuration step in F<config.h>):
493
494 Uid_t who = ...;
495 printf("who = %d\n", who); /* BAD */
496
497The problem here is that Uid_t might be not only not C<int>-wide but it
498might also be unsigned, in which case large uids would be printed as
499negative values.
500
501There is no simple solution to this because of printf()'s limited
502intelligence, but for many types the right format is available as with
503either 'f' or '_f' suffix, for example:
504
505 IVdf /* IV in decimal */
506 UVxf /* UV is hexadecimal */
507
508 printf("i = %"IVdf"\n", i); /* The IVdf is a string constant. */
509
510 Uid_t_f /* Uid_t in decimal */
511
512 printf("who = %"Uid_t_f"\n", who);
513
514Or you can try casting to a "wide enough" type:
515
516 printf("i = %"IVdf"\n", (IV)something_very_small_and_signed);
517
9dd1a77d
KW
518See L<perlguts/Formatted Printing of Size_t and SSize_t> for how to
519print those.
520
04c692a8
DR
521Also remember that the C<%p> format really does require a void pointer:
522
523 U8* p = ...;
524 printf("p = %p\n", (void*)p);
525
526The gcc option C<-Wformat> scans for such problems.
527
528=item *
529
530Blindly using variadic macros
531
532gcc has had them for a while with its own syntax, and C99 brought them
9b22382a 533with a standardized syntax. Don't use the former, and use the latter
04c692a8
DR
534only if the HAS_C99_VARIADIC_MACROS is defined.
535
536=item *
537
538Blindly passing va_list
539
540Not all platforms support passing va_list to further varargs (stdarg)
9b22382a 541functions. The right thing to do is to copy the va_list using the
04c692a8
DR
542Perl_va_copy() if the NEED_VA_COPY is defined.
543
544=item *
545
546Using gcc statement expressions
547
548 val = ({...;...;...}); /* BAD */
549
9b22382a 550While a nice extension, it's not portable. The Perl code does
04c692a8
DR
551admittedly use them if available to gain some extra speed (essentially
552as a funky form of inlining), but you shouldn't.
553
554=item *
555
556Binding together several statements in a macro
557
558Use the macros STMT_START and STMT_END.
559
560 STMT_START {
561 ...
562 } STMT_END
563
564=item *
565
566Testing for operating systems or versions when should be testing for
567features
568
569 #ifdef __FOONIX__ /* BAD */
570 foo = quux();
571 #endif
572
573Unless you know with 100% certainty that quux() is only ever available
574for the "Foonix" operating system B<and> that is available B<and>
575correctly working for B<all> past, present, B<and> future versions of
9b22382a 576"Foonix", the above is very wrong. This is more correct (though still
04c692a8
DR
577not perfect, because the below is a compile-time check):
578
579 #ifdef HAS_QUUX
580 foo = quux();
581 #endif
582
583How does the HAS_QUUX become defined where it needs to be? Well, if
584Foonix happens to be Unixy enough to be able to run the Configure
585script, and Configure has been taught about detecting and testing
9b22382a 586quux(), the HAS_QUUX will be correctly defined. In other platforms, the
04c692a8
DR
587corresponding configuration step will hopefully do the same.
588
589In a pinch, if you cannot wait for Configure to be educated, or if you
590have a good hunch of where quux() might be available, you can
591temporarily try the following:
592
593 #if (defined(__FOONIX__) || defined(__BARNIX__))
594 # define HAS_QUUX
595 #endif
596
597 ...
598
599 #ifdef HAS_QUUX
600 foo = quux();
601 #endif
602
603But in any case, try to keep the features and operating systems
604separate.
605
b39b5b0c
JH
606A good resource on the predefined macros for various operating
607systems, compilers, and so forth is
608L<http://sourceforge.net/p/predef/wiki/Home/>
609
38f18a30
KW
610=item *
611
612Assuming the contents of static memory pointed to by the return values
613of Perl wrappers for C library functions doesn't change. Many C library
614functions return pointers to static storage that can be overwritten by
615subsequent calls to the same or related functions. Perl has
616light-weight wrappers for some of these functions, and which don't make
617copies of the static memory. A good example is the interface to the
618environment variables that are in effect for the program. Perl has
619C<PerlEnv_getenv> to get values from the environment. But the return is
620a pointer to static memory in the C library. If you are using the value
621to immediately test for something, that's fine, but if you save the
622value and expect it to be unchanged by later processing, you would be
623wrong, but perhaps you wouldn't know it because different C library
624implementations behave differently, and the one on the platform you're
625testing on might work for your situation. But on some platforms, a
626subsequent call to C<PerlEnv_getenv> or related function WILL overwrite
627the memory that your first call points to. This has led to some
628hard-to-debug problems. Do a L<perlapi/savepv> to make a copy, thus
629avoiding these problems. You will have to free the copy when you're
630done to avoid memory leaks. If you don't have control over when it gets
631freed, you'll need to make the copy in a mortal scalar, like so:
632
633 if ((s = PerlEnv_getenv("foo") == NULL) {
634 ... /* handle NULL case */
635 }
636 else {
637 s = SvPVX(sv_2mortal(newSVpv(s, 0)));
638 }
639
640The above example works only if C<"s"> is C<NUL>-terminated; otherwise
641you have to pass its length to C<newSVpv>.
642
04c692a8
DR
643=back
644
645=head2 Problematic System Interfaces
646
647=over 4
648
649=item *
650
9b22382a
FC
651malloc(0), realloc(0), calloc(0, 0) are non-portable. To be portable
652allocate at least one byte. (In general you should rarely need to work
04c692a8
DR
653at this low level, but instead use the various malloc wrappers.)
654
4059ba87
AC
655=item *
656
657snprintf() - the return type is unportable. Use my_snprintf() instead.
658
04c692a8
DR
659=back
660
661=head2 Security problems
662
663Last but not least, here are various tips for safer coding.
bbc89b61 664See also L<perlclib> for libc/stdio replacements one should use.
04c692a8
DR
665
666=over 4
667
668=item *
669
670Do not use gets()
671
9b22382a 672Or we will publicly ridicule you. Seriously.
04c692a8
DR
673
674=item *
675
bbc89b61
JH
676Do not use tmpfile()
677
678Use mkstemp() instead.
679
680=item *
681
04c692a8
DR
682Do not use strcpy() or strcat() or strncpy() or strncat()
683
684Use my_strlcpy() and my_strlcat() instead: they either use the native
685implementation, or Perl's own implementation (borrowed from the public
686domain implementation of INN).
687
688=item *
689
690Do not use sprintf() or vsprintf()
691
4059ba87
AC
692If you really want just plain byte strings, use my_snprintf() and
693my_vsnprintf() instead, which will try to use snprintf() and
694vsnprintf() if those safer APIs are available. If you want something
6bfe0388
KW
695fancier than a plain byte string, use
696L<C<Perl_form>()|perlapi/form> or SVs and
697L<C<Perl_sv_catpvf()>|perlapi/sv_catpvf>.
698
4059ba87
AC
699Note that glibc C<printf()>, C<sprintf()>, etc. are buggy before glibc
700version 2.17. They won't allow a C<%.s> format with a precision to
701create a string that isn't valid UTF-8 if the current underlying locale
702of the program is UTF-8. What happens is that the C<%s> and its operand are
703simply skipped without any notice.
704L<https://sourceware.org/bugzilla/show_bug.cgi?id=6530>.
705
c98823ff
JH
706=item *
707
708Do not use atoi()
709
22ff3130 710Use grok_atoUV() instead. atoi() has ill-defined behavior on overflows,
c98823ff 711and cannot be used for incremental parsing. It is also affected by locale,
338aa8b0
JH
712which is bad.
713
714=item *
715
716Do not use strtol() or strtoul()
717
22ff3130 718Use grok_atoUV() instead. strtol() or strtoul() (or their IV/UV-friendly
338aa8b0
JH
719macro disguises, Strtol() and Strtoul(), or Atol() and Atoul() are
720affected by locale, which is bad.
c98823ff 721
04c692a8
DR
722=back
723
724=head1 DEBUGGING
725
726You can compile a special debugging version of Perl, which allows you
727to use the C<-D> option of Perl to tell more about what Perl is doing.
728But sometimes there is no alternative than to dive in with a debugger,
729either to see the stack trace of a core dump (very useful in a bug
730report), or trying to figure out what went wrong before the core dump
731happened, or how did we end up having wrong or unexpected results.
732
733=head2 Poking at Perl
734
735To really poke around with Perl, you'll probably want to build Perl for
736debugging, like this:
737
f075db89 738 ./Configure -d -DDEBUGGING
04c692a8
DR
739 make
740
f075db89
DM
741C<-DDEBUGGING> turns on the C compiler's C<-g> flag to have it produce
742debugging information which will allow us to step through a running
743program, and to see in which C function we are at (without the debugging
744information we might see only the numerical addresses of the functions,
745which is not very helpful). It will also turn on the C<DEBUGGING>
746compilation symbol which enables all the internal debugging code in Perl.
747There are a whole bunch of things you can debug with this: L<perlrun>
748lists them all, and the best way to find out about them is to play about
749with them. The most useful options are probably
04c692a8
DR
750
751 l Context (loop) stack processing
f075db89 752 s Stack snapshots (with v, displays all stacks)
04c692a8
DR
753 t Trace execution
754 o Method and overloading resolution
755 c String/numeric conversions
756
f075db89
DM
757For example
758
759 $ perl -Dst -e '$a + 1'
760 ....
761 (-e:1) gvsv(main::a)
762 => UNDEF
763 (-e:1) const(IV(1))
764 => UNDEF IV(1)
765 (-e:1) add
766 => NV(1)
767
768
769Some of the functionality of the debugging code can be achieved with a
770non-debugging perl by using XS modules:
04c692a8
DR
771
772 -Dr => use re 'debug'
773 -Dx => use O 'Debug'
774
775=head2 Using a source-level debugger
776
777If the debugging output of C<-D> doesn't help you, it's time to step
778through perl's execution with a source-level debugger.
779
780=over 3
781
782=item *
783
784We'll use C<gdb> for our examples here; the principles will apply to
785any debugger (many vendors call their debugger C<dbx>), but check the
786manual of the one you're using.
787
788=back
789
790To fire up the debugger, type
791
792 gdb ./perl
793
794Or if you have a core dump:
795
796 gdb ./perl core
797
798You'll want to do that in your Perl source tree so the debugger can
9b22382a 799read the source code. You should see the copyright message, followed by
04c692a8
DR
800the prompt.
801
802 (gdb)
803
804C<help> will get you into the documentation, but here are the most
805useful commands:
806
807=over 3
808
809=item * run [args]
810
811Run the program with the given arguments.
812
813=item * break function_name
814
815=item * break source.c:xxx
816
817Tells the debugger that we'll want to pause execution when we reach
818either the named function (but see L<perlguts/Internal Functions>!) or
819the given line in the named source file.
820
821=item * step
822
823Steps through the program a line at a time.
824
825=item * next
826
827Steps through the program a line at a time, without descending into
828functions.
829
830=item * continue
831
832Run until the next breakpoint.
833
834=item * finish
835
836Run until the end of the current function, then stop again.
837
838=item * 'enter'
839
840Just pressing Enter will do the most recent operation again - it's a
841blessing when stepping through miles of source code.
842
8b029fdf
MH
843=item * ptype
844
845Prints the C definition of the argument given.
846
847 (gdb) ptype PL_op
848 type = struct op {
849 OP *op_next;
86cd3a13 850 OP *op_sibparent;
8b029fdf
MH
851 OP *(*op_ppaddr)(void);
852 PADOFFSET op_targ;
853 unsigned int op_type : 9;
854 unsigned int op_opt : 1;
855 unsigned int op_slabbed : 1;
856 unsigned int op_savefree : 1;
857 unsigned int op_static : 1;
858 unsigned int op_folded : 1;
859 unsigned int op_spare : 2;
860 U8 op_flags;
861 U8 op_private;
862 } *
863
04c692a8
DR
864=item * print
865
9b22382a 866Execute the given C code and print its results. B<WARNING>: Perl makes
04c692a8 867heavy use of macros, and F<gdb> does not necessarily support macros
9b22382a 868(see later L</"gdb macro support">). You'll have to substitute them
04c692a8
DR
869yourself, or to invoke cpp on the source code files (see L</"The .i
870Targets">) So, for instance, you can't say
871
872 print SvPV_nolen(sv)
873
874but you have to say
875
876 print Perl_sv_2pv_nolen(sv)
877
878=back
879
880You may find it helpful to have a "macro dictionary", which you can
9b22382a 881produce by saying C<cpp -dM perl.c | sort>. Even then, F<cpp> won't
04c692a8
DR
882recursively apply those macros for you.
883
884=head2 gdb macro support
885
886Recent versions of F<gdb> have fairly good macro support, but in order
887to use it you'll need to compile perl with macro definitions included
9b22382a
FC
888in the debugging information. Using F<gcc> version 3.1, this means
889configuring with C<-Doptimize=-g3>. Other compilers might use a
04c692a8
DR
890different switch (if they support debugging macros at all).
891
892=head2 Dumping Perl Data Structures
893
894One way to get around this macro hell is to use the dumping functions
895in F<dump.c>; these work a little like an internal
896L<Devel::Peek|Devel::Peek>, but they also cover OPs and other
9b22382a 897structures that you can't get at from Perl. Let's take an example.
04c692a8 898We'll use the C<$a = $b + $c> we used before, but give it a bit of
9b22382a 899context: C<$b = "6XXXX"; $c = 2.3;>. Where's a good place to stop and
04c692a8
DR
900poke around?
901
902What about C<pp_add>, the function we examined earlier to implement the
903C<+> operator:
904
905 (gdb) break Perl_pp_add
906 Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.
907
908Notice we use C<Perl_pp_add> and not C<pp_add> - see
9b22382a 909L<perlguts/Internal Functions>. With the breakpoint in place, we can
04c692a8
DR
910run our program:
911
912 (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'
913
914Lots of junk will go past as gdb reads in the relevant source files and
915libraries, and then:
916
917 Breakpoint 1, Perl_pp_add () at pp_hot.c:309
72876cce 918 1396 dSP; dATARGET; bool useleft; SV *svl, *svr;
04c692a8
DR
919 (gdb) step
920 311 dPOPTOPnnrl_ul;
921 (gdb)
922
923We looked at this bit of code before, and we said that
924C<dPOPTOPnnrl_ul> arranges for two C<NV>s to be placed into C<left> and
925C<right> - let's slightly expand it:
926
927 #define dPOPTOPnnrl_ul NV right = POPn; \
928 SV *leftsv = TOPs; \
929 NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0
930
931C<POPn> takes the SV from the top of the stack and obtains its NV
932either directly (if C<SvNOK> is set) or by calling the C<sv_2nv>
9b22382a
FC
933function. C<TOPs> takes the next SV from the top of the stack - yes,
934C<POPn> uses C<TOPs> - but doesn't remove it. We then use C<SvNV> to
04c692a8
DR
935get the NV from C<leftsv> in the same way as before - yes, C<POPn> uses
936C<SvNV>.
937
938Since we don't have an NV for C<$b>, we'll have to use C<sv_2nv> to
9b22382a 939convert it. If we step again, we'll find ourselves there:
04c692a8 940
8b029fdf 941 (gdb) step
04c692a8
DR
942 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
943 1669 if (!sv)
944 (gdb)
945
946We can now use C<Perl_sv_dump> to investigate the SV:
947
8b029fdf 948 (gdb) print Perl_sv_dump(sv)
04c692a8
DR
949 SV = PV(0xa057cc0) at 0xa0675d0
950 REFCNT = 1
951 FLAGS = (POK,pPOK)
952 PV = 0xa06a510 "6XXXX"\0
953 CUR = 5
954 LEN = 6
955 $1 = void
956
957We know we're going to get C<6> from this, so let's finish the
958subroutine:
959
960 (gdb) finish
961 Run till exit from #0 Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
962 0x462669 in Perl_pp_add () at pp_hot.c:311
963 311 dPOPTOPnnrl_ul;
964
965We can also dump out this op: the current op is always stored in
9b22382a 966C<PL_op>, and we can dump it with C<Perl_op_dump>. This'll give us
903b1101 967similar output to CPAN module B::Debug.
04c692a8 968
8b029fdf 969 (gdb) print Perl_op_dump(PL_op)
04c692a8
DR
970 {
971 13 TYPE = add ===> 14
972 TARG = 1
973 FLAGS = (SCALAR,KIDS)
974 {
975 TYPE = null ===> (12)
976 (was rv2sv)
977 FLAGS = (SCALAR,KIDS)
978 {
979 11 TYPE = gvsv ===> 12
980 FLAGS = (SCALAR)
981 GV = main::b
982 }
983 }
984
985# finish this later #
986
8b029fdf
MH
987=head2 Using gdb to look at specific parts of a program
988
73013070
SF
989With the example above, you knew to look for C<Perl_pp_add>, but what if
990there were multiple calls to it all over the place, or you didn't know what
8b029fdf
MH
991the op was you were looking for?
992
73013070 993One way to do this is to inject a rare call somewhere near what you're looking
9b22382a 994for. For example, you could add C<study> before your method:
8b029fdf
MH
995
996 study;
997
998And in gdb do:
999
1000 (gdb) break Perl_pp_study
1001
9b22382a 1002And then step until you hit what you're
73013070 1003looking for. This works well in a loop
8b029fdf
MH
1004if you want to only break at certain iterations:
1005
1006 for my $c (1..100) {
1007 study if $c == 50;
1008 }
1009
1010=head2 Using gdb to look at what the parser/lexer are doing
1011
73013070 1012If you want to see what perl is doing when parsing/lexing your code, you can
72b22e55 1013use C<BEGIN {}>:
8b029fdf
MH
1014
1015 print "Before\n";
1016 BEGIN { study; }
1017 print "After\n";
1018
1019And in gdb:
1020
1021 (gdb) break Perl_pp_study
1022
1023If you want to see what the parser/lexer is doing inside of C<if> blocks and
1024the like you need to be a little trickier:
1025
73013070 1026 if ($a && $b && do { BEGIN { study } 1 } && $c) { ... }
8b029fdf 1027
04c692a8
DR
1028=head1 SOURCE CODE STATIC ANALYSIS
1029
1030Various tools exist for analysing C source code B<statically>, as
9b22382a 1031opposed to B<dynamically>, that is, without executing the code. It is
04c692a8
DR
1032possible to detect resource leaks, undefined behaviour, type
1033mismatches, portability problems, code paths that would cause illegal
1034memory accesses, and other similar problems by just parsing the C code
1035and looking at the resulting graph, what does it tell about the
9b22382a 1036execution and data flows. As a matter of fact, this is exactly how C
04c692a8
DR
1037compilers know to give warnings about dubious code.
1038
c707756e 1039=head2 lint
04c692a8
DR
1040
1041The good old C code quality inspector, C<lint>, is available in several
1042platforms, but please be aware that there are several different
1043implementations of it by different vendors, which means that the flags
1044are not identical across different platforms.
1045
c707756e 1046There is a C<lint> target in Makefile, but you may have to
04c692a8
DR
1047diddle with the flags (see above).
1048
1049=head2 Coverity
1050
4b05bc8e 1051Coverity (L<http://www.coverity.com/>) is a product similar to lint and as
04c692a8
DR
1052a testbed for their product they periodically check several open source
1053projects, and they give out accounts to open source developers to the
1054defect databases.
1055
d3c1eddb
JH
1056There is Coverity setup for the perl5 project:
1057L<https://scan.coverity.com/projects/perl5>
1058
a72f2680 1059=head2 HP-UX cadvise (Code Advisor)
65c4791f
JH
1060
1061HP has a C/C++ static analyzer product for HP-UX caller Code Advisor.
1062(Link not given here because the URL is horribly long and seems horribly
1063unstable; use the search engine of your choice to find it.) The use of
1064the C<cadvise_cc> recipe with C<Configure ... -Dcc=./cadvise_cc>
1065(see cadvise "User Guide") is recommended; as is the use of C<+wall>.
1066
04c692a8
DR
1067=head2 cpd (cut-and-paste detector)
1068
9b22382a 1069The cpd tool detects cut-and-paste coding. If one instance of the
04c692a8 1070cut-and-pasted code changes, all the other spots should probably be
9b22382a 1071changed, too. Therefore such code should probably be turned into a
04c692a8
DR
1072subroutine or a macro.
1073
4b05bc8e
KW
1074cpd (L<http://pmd.sourceforge.net/cpd.html>) is part of the pmd project
1075(L<http://pmd.sourceforge.net/>). pmd was originally written for static
04c692a8
DR
1076analysis of Java code, but later the cpd part of it was extended to
1077parse also C and C++.
1078
1079Download the pmd-bin-X.Y.zip () from the SourceForge site, extract the
1080pmd-X.Y.jar from it, and then run that on source code thusly:
1081
0cbf2b31
FC
1082 java -cp pmd-X.Y.jar net.sourceforge.pmd.cpd.CPD \
1083 --minimum-tokens 100 --files /some/where/src --language c > cpd.txt
04c692a8
DR
1084
1085You may run into memory limits, in which case you should use the -Xmx
1086option:
1087
1088 java -Xmx512M ...
1089
1090=head2 gcc warnings
1091
1092Though much can be written about the inconsistency and coverage
1093problems of gcc warnings (like C<-Wall> not meaning "all the warnings",
1094or some common portability problems not being covered by C<-Wall>, or
1095C<-ansi> and C<-pedantic> both being a poorly defined collection of
1096warnings, and so forth), gcc is still a useful tool in keeping our
1097coding nose clean.
1098
1099The C<-Wall> is by default on.
1100
1101The C<-ansi> (and its sidekick, C<-pedantic>) would be nice to be on
1102always, but unfortunately they are not safe on all platforms, they can
1103for example cause fatal conflicts with the system headers (Solaris
9b22382a 1104being a prime example). If Configure C<-Dgccansipedantic> is used, the
04c692a8
DR
1105C<cflags> frontend selects C<-ansi -pedantic> for the platforms where
1106they are known to be safe.
1107
2884c977 1108The following extra flags are added:
04c692a8
DR
1109
1110=over 4
1111
1112=item *
1113
1114C<-Wendif-labels>
1115
1116=item *
1117
1118C<-Wextra>
1119
1120=item *
1121
2884c977
DIM
1122C<-Wc++-compat>
1123
1124=item *
1125
1126C<-Wwrite-strings>
1127
1128=item *
1129
1130C<-Werror=declaration-after-statement>
04c692a8 1131
5997475b
DIM
1132=item *
1133
1134C<-Werror=pointer-arith>
1135
04c692a8
DR
1136=back
1137
1138The following flags would be nice to have but they would first need
1139their own Augean stablemaster:
1140
1141=over 4
1142
1143=item *
1144
04c692a8
DR
1145C<-Wshadow>
1146
1147=item *
1148
1149C<-Wstrict-prototypes>
1150
1151=back
1152
1153The C<-Wtraditional> is another example of the annoying tendency of gcc
1154to bundle a lot of warnings under one switch (it would be impossible to
1155deploy in practice because it would complain a lot) but it does contain
1156some warnings that would be beneficial to have available on their own,
1157such as the warning about string constants inside macros containing the
1158macro arguments: this behaved differently pre-ANSI than it does in
1159ANSI, and some C compilers are still in transition, AIX being an
1160example.
1161
1162=head2 Warnings of other C compilers
1163
1164Other C compilers (yes, there B<are> other C compilers than gcc) often
1165have their "strict ANSI" or "strict ANSI with some portability
1166extensions" modes on, like for example the Sun Workshop has its C<-Xa>
1167mode on (though implicitly), or the DEC (these days, HP...) has its
1168C<-std1> mode on.
1169
1170=head1 MEMORY DEBUGGERS
1171
d1fd4856
VP
1172B<NOTE 1>: Running under older memory debuggers such as Purify,
1173valgrind or Third Degree greatly slows down the execution: seconds
9b22382a 1174become minutes, minutes become hours. For example as of Perl 5.8.1, the
04c692a8 1175ext/Encode/t/Unicode.t takes extraordinarily long to complete under
9b22382a
FC
1176e.g. Purify, Third Degree, and valgrind. Under valgrind it takes more
1177than six hours, even on a snappy computer. The said test must be doing
1178something that is quite unfriendly for memory debuggers. If you don't
04c692a8 1179feel like waiting, that you can simply kill away the perl process.
d1fd4856
VP
1180Roughly valgrind slows down execution by factor 10, AddressSanitizer by
1181factor 2.
04c692a8
DR
1182
1183B<NOTE 2>: To minimize the number of memory leak false alarms (see
1184L</PERL_DESTRUCT_LEVEL> for more information), you have to set the
9b22382a 1185environment variable PERL_DESTRUCT_LEVEL to 2. For example, like this:
04c692a8
DR
1186
1187 env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib ...
1188
1189B<NOTE 3>: There are known memory leaks when there are compile-time
1190errors within eval or require, seeing C<S_doeval> in the call stack is
9b22382a 1191a good sign of these. Fixing these leaks is non-trivial, unfortunately,
04c692a8
DR
1192but they must be fixed eventually.
1193
1194B<NOTE 4>: L<DynaLoader> will not clean up after itself completely
1195unless Perl is built with the Configure option
1196C<-Accflags=-DDL_UNLOAD_ALL_AT_EXIT>.
1197
04c692a8
DR
1198=head2 valgrind
1199
d1fd4856 1200The valgrind tool can be used to find out both memory leaks and illegal
9b22382a 1201heap memory accesses. As of version 3.3.0, Valgrind only supports Linux
0263e49a 1202on x86, x86-64 and PowerPC and Darwin (OS X) on x86 and x86-64. The
d1fd4856 1203special "test.valgrind" target can be used to run the tests under
9b22382a 1204valgrind. Found errors and memory leaks are logged in files named
037ab3f1
MH
1205F<testfile.valgrind> and by default output is displayed inline.
1206
1207Example usage:
1208
1209 make test.valgrind
1210
1211Since valgrind adds significant overhead, tests will take much longer to
1212run. The valgrind tests support being run in parallel to help with this:
1213
1214 TEST_JOBS=9 make test.valgrind
1215
1216Note that the above two invocations will be very verbose as reachable
1217memory and leak-checking is enabled by default. If you want to just see
1218pure errors, try:
73013070 1219
037ab3f1
MH
1220 VG_OPTS='-q --leak-check=no --show-reachable=no' TEST_JOBS=9 \
1221 make test.valgrind
04c692a8
DR
1222
1223Valgrind also provides a cachegrind tool, invoked on perl as:
1224
1225 VG_OPTS=--tool=cachegrind make test.valgrind
1226
1227As system libraries (most notably glibc) are also triggering errors,
9b22382a 1228valgrind allows to suppress such errors using suppression files. The
04c692a8 1229default suppression file that comes with valgrind already catches a lot
9b22382a 1230of them. Some additional suppressions are defined in F<t/perl.supp>.
04c692a8
DR
1231
1232To get valgrind and for more information see
1233
0061d4fa 1234 http://valgrind.org/
04c692a8 1235
81c3bbe7
RU
1236=head2 AddressSanitizer
1237
4dd56148 1238AddressSanitizer is a clang and gcc extension, included in clang since
9b22382a 1239v3.1 and gcc since v4.8. It checks illegal heap pointers, global
4dd56148
NC
1240pointers, stack pointers and use after free errors, and is fast enough
1241that you can easily compile your debugging or optimized perl with it.
9b22382a 1242It does not check memory leaks though. AddressSanitizer is available
4dd56148 1243for Linux, Mac OS X and soon on Windows.
81c3bbe7 1244
8a64fbaa
VP
1245To build perl with AddressSanitizer, your Configure invocation should
1246look like:
81c3bbe7 1247
e8596d90
VP
1248 sh Configure -des -Dcc=clang \
1249 -Accflags=-faddress-sanitizer -Aldflags=-faddress-sanitizer \
1250 -Alddlflags=-shared\ -faddress-sanitizer
81c3bbe7
RU
1251
1252where these arguments mean:
1253
1254=over 4
1255
1256=item * -Dcc=clang
1257
8a64fbaa
VP
1258This should be replaced by the full path to your clang executable if it
1259is not in your path.
81c3bbe7
RU
1260
1261=item * -Accflags=-faddress-sanitizer
1262
8a64fbaa 1263Compile perl and extensions sources with AddressSanitizer.
81c3bbe7
RU
1264
1265=item * -Aldflags=-faddress-sanitizer
1266
8a64fbaa 1267Link the perl executable with AddressSanitizer.
81c3bbe7 1268
e8596d90 1269=item * -Alddlflags=-shared\ -faddress-sanitizer
81c3bbe7 1270
9b22382a 1271Link dynamic extensions with AddressSanitizer. You must manually
e8596d90
VP
1272specify C<-shared> because using C<-Alddlflags=-shared> will prevent
1273Configure from setting a default value for C<lddlflags>, which usually
5dfc6e97 1274contains C<-shared> (at least on Linux).
81c3bbe7
RU
1275
1276=back
1277
8a64fbaa
VP
1278See also
1279L<http://code.google.com/p/address-sanitizer/wiki/AddressSanitizer>.
81c3bbe7
RU
1280
1281
04c692a8
DR
1282=head1 PROFILING
1283
1284Depending on your platform there are various ways of profiling Perl.
1285
1286There are two commonly used techniques of profiling executables:
1287I<statistical time-sampling> and I<basic-block counting>.
1288
1289The first method takes periodically samples of the CPU program counter,
1290and since the program counter can be correlated with the code generated
1291for functions, we get a statistical view of in which functions the
9b22382a 1292program is spending its time. The caveats are that very small/fast
04c692a8
DR
1293functions have lower probability of showing up in the profile, and that
1294periodically interrupting the program (this is usually done rather
1295frequently, in the scale of milliseconds) imposes an additional
9b22382a 1296overhead that may skew the results. The first problem can be alleviated
04c692a8
DR
1297by running the code for longer (in general this is a good idea for
1298profiling), the second problem is usually kept in guard by the
1299profiling tools themselves.
1300
1301The second method divides up the generated code into I<basic blocks>.
1302Basic blocks are sections of code that are entered only in the
9b22382a
FC
1303beginning and exited only at the end. For example, a conditional jump
1304starts a basic block. Basic block profiling usually works by
04c692a8 1305I<instrumenting> the code by adding I<enter basic block #nnnn>
9b22382a
FC
1306book-keeping code to the generated code. During the execution of the
1307code the basic block counters are then updated appropriately. The
04c692a8
DR
1308caveat is that the added extra code can skew the results: again, the
1309profiling tools usually try to factor their own effects out of the
1310results.
1311
1312=head2 Gprof Profiling
1313
e2aed43d 1314I<gprof> is a profiling tool available in many Unix platforms which
9b22382a
FC
1315uses I<statistical time-sampling>. You can build a profiled version of
1316F<perl> by compiling using gcc with the flag C<-pg>. Either edit
1317F<config.sh> or re-run F<Configure>. Running the profiled version of
e2aed43d
NC
1318Perl will create an output file called F<gmon.out> which contains the
1319profiling data collected during the execution.
04c692a8 1320
e2aed43d
NC
1321quick hint:
1322
1323 $ sh Configure -des -Dusedevel -Accflags='-pg' \
1324 -Aldflags='-pg' -Alddlflags='-pg -shared' \
1325 && make perl
1326 $ ./perl ... # creates gmon.out in current directory
1327 $ gprof ./perl > out
1328 $ less out
1329
1330(you probably need to add C<-shared> to the <-Alddlflags> line until RT
1331#118199 is resolved)
04c692a8 1332
e2aed43d
NC
1333The F<gprof> tool can then display the collected data in various ways.
1334Usually F<gprof> understands the following options:
04c692a8
DR
1335
1336=over 4
1337
1338=item * -a
1339
1340Suppress statically defined functions from the profile.
1341
1342=item * -b
1343
1344Suppress the verbose descriptions in the profile.
1345
1346=item * -e routine
1347
1348Exclude the given routine and its descendants from the profile.
1349
1350=item * -f routine
1351
1352Display only the given routine and its descendants in the profile.
1353
1354=item * -s
1355
1356Generate a summary file called F<gmon.sum> which then may be given to
1357subsequent gprof runs to accumulate data over several runs.
1358
1359=item * -z
1360
1361Display routines that have zero usage.
1362
1363=back
1364
1365For more detailed explanation of the available commands and output
e2aed43d 1366formats, see your own local documentation of F<gprof>.
04c692a8 1367
e2aed43d 1368=head2 GCC gcov Profiling
04c692a8 1369
e2aed43d
NC
1370I<basic block profiling> is officially available in gcc 3.0 and later.
1371You can build a profiled version of F<perl> by compiling using gcc with
9b22382a 1372the flags C<-fprofile-arcs -ftest-coverage>. Either edit F<config.sh>
e2aed43d 1373or re-run F<Configure>.
04c692a8 1374
e2aed43d 1375quick hint:
04c692a8 1376
e2aed43d
NC
1377 $ sh Configure -des -Dusedevel -Doptimize='-g' \
1378 -Accflags='-fprofile-arcs -ftest-coverage' \
1379 -Aldflags='-fprofile-arcs -ftest-coverage' \
1380 -Alddlflags='-fprofile-arcs -ftest-coverage -shared' \
1381 && make perl
1382 $ rm -f regexec.c.gcov regexec.gcda
1383 $ ./perl ...
1384 $ gcov regexec.c
1385 $ less regexec.c.gcov
04c692a8 1386
e2aed43d
NC
1387(you probably need to add C<-shared> to the <-Alddlflags> line until RT
1388#118199 is resolved)
04c692a8
DR
1389
1390Running the profiled version of Perl will cause profile output to be
9b22382a 1391generated. For each source file an accompanying F<.gcda> file will be
04c692a8
DR
1392created.
1393
e2aed43d 1394To display the results you use the I<gcov> utility (which should be
9b22382a 1395installed if you have gcc 3.0 or newer installed). F<gcov> is run on
04c692a8
DR
1396source code files, like this
1397
1398 gcov sv.c
1399
9b22382a 1400which will cause F<sv.c.gcov> to be created. The F<.gcov> files contain
04c692a8 1401the source code annotated with relative frequencies of execution
9b22382a 1402indicated by "#" markers. If you want to generate F<.gcov> files for
6f134219
NC
1403all profiled object files, you can run something like this:
1404
1405 for file in `find . -name \*.gcno`
1406 do sh -c "cd `dirname $file` && gcov `basename $file .gcno`"
1407 done
04c692a8
DR
1408
1409Useful options of F<gcov> include C<-b> which will summarise the basic
1410block, branch, and function call coverage, and C<-c> which instead of
9b22382a 1411relative frequencies will use the actual counts. For more information
04c692a8 1412on the use of F<gcov> and basic block profiling with gcc, see the
9b22382a 1413latest GNU CC manual. As of gcc 4.8, this is at
e2aed43d 1414L<http://gcc.gnu.org/onlinedocs/gcc/Gcov-Intro.html#Gcov-Intro>
04c692a8
DR
1415
1416=head1 MISCELLANEOUS TRICKS
1417
1418=head2 PERL_DESTRUCT_LEVEL
1419
1420If you want to run any of the tests yourself manually using e.g.
4dd56148
NC
1421valgrind, please note that by default perl B<does not> explicitly
1422cleanup all the memory it has allocated (such as global memory arenas)
1423but instead lets the exit() of the whole program "take care" of such
1424allocations, also known as "global destruction of objects".
04c692a8
DR
1425
1426There is a way to tell perl to do complete cleanup: set the environment
9b22382a 1427variable PERL_DESTRUCT_LEVEL to a non-zero value. The t/TEST wrapper
04c692a8 1428does set this to 2, and this is what you need to do too, if you don't
f01ecde8 1429want to see the "global leaks": For example, for running under valgrind
04c692a8 1430
a63ef199 1431 env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib t/foo/bar.t
04c692a8
DR
1432
1433(Note: the mod_perl apache module uses also this environment variable
9b22382a
FC
1434for its own purposes and extended its semantics. Refer to the mod_perl
1435documentation for more information. Also, spawned threads do the
04c692a8
DR
1436equivalent of setting this variable to the value 1.)
1437
1438If, at the end of a run you get the message I<N scalars leaked>, you
b4986286
DM
1439can recompile with C<-DDEBUG_LEAKING_SCALARS>,
1440(C<Configure -Accflags=-DDEBUG_LEAKING_SCALARS>), which will cause the
04c692a8 1441addresses of all those leaked SVs to be dumped along with details as to
9b22382a
FC
1442where each SV was originally allocated. This information is also
1443displayed by Devel::Peek. Note that the extra details recorded with
04c692a8 1444each SV increases memory usage, so it shouldn't be used in production
9b22382a 1445environments. It also converts C<new_SV()> from a macro into a real
04c692a8
DR
1446function, so you can use your favourite debugger to discover where
1447those pesky SVs were allocated.
1448
1449If you see that you're leaking memory at runtime, but neither valgrind
1450nor C<-DDEBUG_LEAKING_SCALARS> will find anything, you're probably
1451leaking SVs that are still reachable and will be properly cleaned up
9b22382a
FC
1452during destruction of the interpreter. In such cases, using the C<-Dm>
1453switch can point you to the source of the leak. If the executable was
04c692a8 1454built with C<-DDEBUG_LEAKING_SCALARS>, C<-Dm> will output SV
9b22382a 1455allocations in addition to memory allocations. Each SV allocation has a
04c692a8 1456distinct serial number that will be written on creation and destruction
9b22382a 1457of the SV. So if you're executing the leaking code in a loop, you need
04c692a8 1458to look for SVs that are created, but never destroyed between each
9b22382a 1459cycle. If such an SV is found, set a conditional breakpoint within
04c692a8 1460C<new_SV()> and make it break only when C<PL_sv_serial> is equal to the
9b22382a 1461serial number of the leaking SV. Then you will catch the interpreter in
04c692a8
DR
1462exactly the state where the leaking SV is allocated, which is
1463sufficient in many cases to find the source of the leak.
1464
1465As C<-Dm> is using the PerlIO layer for output, it will by itself
9b22382a 1466allocate quite a bunch of SVs, which are hidden to avoid recursion. You
04c692a8
DR
1467can bypass the PerlIO layer if you use the SV logging provided by
1468C<-DPERL_MEM_LOG> instead.
1469
1470=head2 PERL_MEM_LOG
1471
6fb87544
MH
1472If compiled with C<-DPERL_MEM_LOG> (C<-Accflags=-DPERL_MEM_LOG>), both
1473memory and SV allocations go through logging functions, which is
1474handy for breakpoint setting.
04c692a8 1475
6fb87544
MH
1476Unless C<-DPERL_MEM_LOG_NOIMPL> (C<-Accflags=-DPERL_MEM_LOG_NOIMPL>) is
1477also compiled, the logging functions read $ENV{PERL_MEM_LOG} to
1478determine whether to log the event, and if so how:
04c692a8 1479
a63ef199
SF
1480 $ENV{PERL_MEM_LOG} =~ /m/ Log all memory ops
1481 $ENV{PERL_MEM_LOG} =~ /s/ Log all SV ops
1482 $ENV{PERL_MEM_LOG} =~ /t/ include timestamp in Log
1483 $ENV{PERL_MEM_LOG} =~ /^(\d+)/ write to FD given (default is 2)
04c692a8
DR
1484
1485Memory logging is somewhat similar to C<-Dm> but is independent of
1486C<-DDEBUGGING>, and at a higher level; all uses of Newx(), Renew(), and
1487Safefree() are logged with the caller's source code file and line
9b22382a
FC
1488number (and C function name, if supported by the C compiler). In
1489contrast, C<-Dm> is directly at the point of C<malloc()>. SV logging is
04c692a8
DR
1490similar.
1491
1492Since the logging doesn't use PerlIO, all SV allocations are logged and
9b22382a 1493no extra SV allocations are introduced by enabling the logging. If
04c692a8
DR
1494compiled with C<-DDEBUG_LEAKING_SCALARS>, the serial number for each SV
1495allocation is also logged.
1496
1497=head2 DDD over gdb
1498
1499Those debugging perl with the DDD frontend over gdb may find the
1500following useful:
1501
1502You can extend the data conversion shortcuts menu, so for example you
1503can display an SV's IV value with one click, without doing any typing.
1504To do that simply edit ~/.ddd/init file and add after:
1505
1506 ! Display shortcuts.
1507 Ddd*gdbDisplayShortcuts: \
1508 /t () // Convert to Bin\n\
1509 /d () // Convert to Dec\n\
1510 /x () // Convert to Hex\n\
1511 /o () // Convert to Oct(\n\
1512
1513the following two lines:
1514
1515 ((XPV*) (())->sv_any )->xpv_pv // 2pvx\n\
1516 ((XPVIV*) (())->sv_any )->xiv_iv // 2ivx
1517
1518so now you can do ivx and pvx lookups or you can plug there the sv_peek
1519"conversion":
1520
1521 Perl_sv_peek(my_perl, (SV*)()) // sv_peek
1522
9b22382a 1523(The my_perl is for threaded builds.) Just remember that every line,
04c692a8
DR
1524but the last one, should end with \n\
1525
1526Alternatively edit the init file interactively via: 3rd mouse button ->
1527New Display -> Edit Menu
1528
1529Note: you can define up to 20 conversion shortcuts in the gdb section.
1530
470dd224
JH
1531=head2 C backtrace
1532
0762e42f
JH
1533On some platforms Perl supports retrieving the C level backtrace
1534(similar to what symbolic debuggers like gdb do).
470dd224
JH
1535
1536The backtrace returns the stack trace of the C call frames,
1537with the symbol names (function names), the object names (like "perl"),
1538and if it can, also the source code locations (file:line).
1539
0762e42f
JH
1540The supported platforms are Linux, and OS X (some *BSD might
1541work at least partly, but they have not yet been tested).
1542
1543This feature hasn't been tested with multiple threads, but it will
1544only show the backtrace of the thread doing the backtracing.
470dd224
JH
1545
1546The feature needs to be enabled with C<Configure -Dusecbacktrace>.
1547
0762e42f
JH
1548The C<-Dusecbacktrace> also enables keeping the debug information when
1549compiling/linking (often: C<-g>). Many compilers/linkers do support
1550having both optimization and keeping the debug information. The debug
1551information is needed for the symbol names and the source locations.
1552
1553Static functions might not be visible for the backtrace.
470dd224
JH
1554
1555Source code locations, even if available, can often be missing or
0762e42f
JH
1556misleading if the compiler has e.g. inlined code. Optimizer can
1557make matching the source code and the object code quite challenging.
470dd224
JH
1558
1559=over 4
1560
1561=item Linux
1562
59b3baca 1563You B<must> have the BFD (-lbfd) library installed, otherwise C<perl> will
0762e42f 1564fail to link. The BFD is usually distributed as part of the GNU binutils.
470dd224
JH
1565
1566Summary: C<Configure ... -Dusecbacktrace>
1567and you need C<-lbfd>.
1568
1569=item OS X
1570
0762e42f
JH
1571The source code locations are supported B<only> if you have
1572the Developer Tools installed. (BFD is B<not> needed.)
470dd224
JH
1573
1574Summary: C<Configure ... -Dusecbacktrace>
1575and installing the Developer Tools would be good.
1576
1577=back
1578
1579Optionally, for trying out the feature, you may want to enable
0762e42f
JH
1580automatic dumping of the backtrace just before a warning or croak (die)
1581message is emitted, by adding C<-Accflags=-DUSE_C_BACKTRACE_ON_ERROR>
1582for Configure.
470dd224
JH
1583
1584Unless the above additional feature is enabled, nothing about the
1585backtrace functionality is visible, except for the Perl/XS level.
1586
1587Furthermore, even if you have enabled this feature to be compiled,
1588you need to enable it in runtime with an environment variable:
0762e42f
JH
1589C<PERL_C_BACKTRACE_ON_ERROR=10>. It must be an integer higher
1590than zero, telling the desired frame count.
470dd224
JH
1591
1592Retrieving the backtrace from Perl level (using for example an XS
1593extension) would be much less exciting than one would hope: normally
1594you would see C<runops>, C<entersub>, and not much else. This API is
1595intended to be called B<from within> the Perl implementation, not from
1596Perl level execution.
1597
0762e42f 1598The C API for the backtrace is as follows:
470dd224
JH
1599
1600=over 4
1601
1602=item get_c_backtrace
1603
1604=item free_c_backtrace
1605
1606=item get_c_backtrace_dump
1607
1608=item dump_c_backtrace
1609
1610=back
1611
04c692a8
DR
1612=head2 Poison
1613
1614If you see in a debugger a memory area mysteriously full of 0xABABABAB
1615or 0xEFEFEFEF, you may be seeing the effect of the Poison() macros, see
1616L<perlclib>.
1617
1618=head2 Read-only optrees
1619
9b22382a 1620Under ithreads the optree is read only. If you want to enforce this, to
04c692a8 1621check for write accesses from buggy code, compile with
91fc0422
FC
1622C<-Accflags=-DPERL_DEBUG_READONLY_OPS>
1623to enable code that allocates op memory
4dd56148
NC
1624via C<mmap>, and sets it read-only when it is attached to a subroutine.
1625Any write access to an op results in a C<SIGBUS> and abort.
04c692a8
DR
1626
1627This code is intended for development only, and may not be portable
9b22382a
FC
1628even to all Unix variants. Also, it is an 80% solution, in that it
1629isn't able to make all ops read only. Specifically it does not apply to
4dd56148 1630op slabs belonging to C<BEGIN> blocks.
04c692a8 1631
4dd56148
NC
1632However, as an 80% solution it is still effective, as it has caught
1633bugs in the past.
04c692a8 1634
f789f6a4
FC
1635=head2 When is a bool not a bool?
1636
1637On pre-C99 compilers, C<bool> is defined as equivalent to C<char>.
b59008ae
AC
1638Consequently assignment of any larger type to a C<bool> is unsafe and may be
1639truncated. The C<cBOOL> macro exists to cast it correctly; you may also find
1640that using it is shorter and clearer than writing out the equivalent
1641conditional expression longhand.
f789f6a4
FC
1642
1643On those platforms and compilers where C<bool> really is a boolean (C++,
1644C99), it is easy to forget the cast. You can force C<bool> to be a C<char>
1645by compiling with C<-Accflags=-DPERL_BOOL_AS_CHAR>. You may also wish to
50e4f4d4
CB
1646run C<Configure> with something like
1647
cbc13c3d 1648 -Accflags='-Wconversion -Wno-sign-conversion -Wno-shorten-64-to-32'
50e4f4d4
CB
1649
1650or your compiler's equivalent to make it easier to spot any unsafe truncations
1651that show up.
f789f6a4 1652
b59008ae
AC
1653The C<TRUE> and C<FALSE> macros are available for situations where using them
1654would clarify intent. (But they always just mean the same as the integers 1 and
16550 regardless, so using them isn't compulsory.)
1656
04c692a8
DR
1657=head2 The .i Targets
1658
1659You can expand the macros in a F<foo.c> file by saying
1660
1661 make foo.i
1662
d1fd4856
VP
1663which will expand the macros using cpp. Don't be scared by the
1664results.
04c692a8
DR
1665
1666=head1 AUTHOR
1667
1668This document was originally written by Nathan Torkington, and is
1669maintained by the perl5-porters mailing list.