1PERLHACKTIPS(1)        Perl Programmers Reference Guide        PERLHACKTIPS(1)
2
3
4

NAME

6       perlhacktips - Tips for Perl core C code hacking
7

DESCRIPTION

9       This document will help you learn the best way to go about hacking on
10       the Perl core C code.  It covers common problems, debugging, profiling,
11       and more.
12
13       If you haven't read perlhack and perlhacktut yet, you might want to do
14       that first.
15

COMMON PROBLEMS

17       Perl source plays by ANSI C89 rules: no C99 (or C++) extensions.  You
18       don't care about some particular platform having broken Perl? I hear
19       there is still a strong demand for J2EE programmers.
20
21   Perl environment problems
22       •   Not compiling with threading
23
24           Compiling with threading (-Duseithreads) completely rewrites the
25           function prototypes of Perl.  You better try your changes with
26           that.  Related to this is the difference between "Perl_-less" and
27           "Perl_-ly" APIs, for example:
28
29             Perl_sv_setiv(aTHX_ ...);
30             sv_setiv(...);
31
32           The first one explicitly passes in the context, which is needed for
33           e.g. threaded builds.  The second one does that implicitly; do not
34           get them mixed.  If you are not passing in a aTHX_, you will need
35           to do a dTHX as the first thing in the function.
36
37           See "How multiple interpreters and concurrency are supported" in
38           perlguts for further discussion about context.
39
40       •   Not compiling with -DDEBUGGING
41
42           The DEBUGGING define exposes more code to the compiler, therefore
43           more ways for things to go wrong.  You should try it.
44
45       •   Introducing (non-read-only) globals
46
47           Do not introduce any modifiable globals, truly global or file
48           static.  They are bad form and complicate multithreading and other
49           forms of concurrency.  The right way is to introduce them as new
50           interpreter variables, see intrpvar.h (at the very end for binary
51           compatibility).
52
53           Introducing read-only (const) globals is okay, as long as you
54           verify with e.g. "nm libperl.a|egrep -v ' [TURtr] '" (if your "nm"
55           has BSD-style output) that the data you added really is read-only.
56           (If it is, it shouldn't show up in the output of that command.)
57
58           If you want to have static strings, make them constant:
59
60             static const char etc[] = "...";
61
62           If you want to have arrays of constant strings, note carefully the
63           right combination of "const"s:
64
65               static const char * const yippee[] =
66                   {"hi", "ho", "silver"};
67
68       •   Not exporting your new function
69
70           Some platforms (Win32, AIX, VMS, OS/2, to name a few) require any
71           function that is part of the public API (the shared Perl library)
72           to be explicitly marked as exported.  See the discussion about
73           embed.pl in perlguts.
74
75       •   Exporting your new function
76
77           The new shiny result of either genuine new functionality or your
78           arduous refactoring is now ready and correctly exported.  So what
79           could possibly go wrong?
80
81           Maybe simply that your function did not need to be exported in the
82           first place.  Perl has a long and not so glorious history of
83           exporting functions that it should not have.
84
85           If the function is used only inside one source code file, make it
86           static.  See the discussion about embed.pl in perlguts.
87
88           If the function is used across several files, but intended only for
89           Perl's internal use (and this should be the common case), do not
90           export it to the public API.  See the discussion about embed.pl in
91           perlguts.
92
93   Portability problems
94       The following are common causes of compilation and/or execution
95       failures, not common to Perl as such.  The C FAQ is good bedtime
96       reading.  Please test your changes with as many C compilers and
97       platforms as possible; we will, anyway, and it's nice to save oneself
98       from public embarrassment.
99
100       If using gcc, you can add the "-std=c89" option which will hopefully
101       catch most of these unportabilities.  (However it might also catch
102       incompatibilities in your system's header files.)
103
104       Use the Configure "-Dgccansipedantic" flag to enable the gcc "-ansi
105       -pedantic" flags which enforce stricter ANSI rules.
106
107       If using the "gcc -Wall" note that not all the possible warnings (like
108       "-Wuninitialized") are given unless you also compile with "-O".
109
110       Note that if using gcc, starting from Perl 5.9.5 the Perl core source
111       code files (the ones at the top level of the source code distribution,
112       but not e.g. the extensions under ext/) are automatically compiled with
113       as many as possible of the "-std=c89", "-ansi", "-pedantic", and a
114       selection of "-W" flags (see cflags.SH).
115
116       Also study perlport carefully to avoid any bad assumptions about the
117       operating system, filesystems, character set, and so forth.
118
119       You may once in a while try a "make microperl" to see whether we can
120       still compile Perl with just the bare minimum of interfaces.  (See
121       README.micro.)
122
123       Do not assume an operating system indicates a certain compiler.
124
125       •   Casting pointers to integers or casting integers to pointers
126
127               void castaway(U8* p)
128               {
129                 IV i = p;
130
131           or
132
133               void castaway(U8* p)
134               {
135                 IV i = (IV)p;
136
137           Both are bad, and broken, and unportable.  Use the PTR2IV() macro
138           that does it right.  (Likewise, there are PTR2UV(), PTR2NV(),
139           INT2PTR(), and NUM2PTR().)
140
141       •   Casting between function pointers and data pointers
142
143           Technically speaking casting between function pointers and data
144           pointers is unportable and undefined, but practically speaking it
145           seems to work, but you should use the FPTR2DPTR() and DPTR2FPTR()
146           macros.  Sometimes you can also play games with unions.
147
148       •   Assuming sizeof(int) == sizeof(long)
149
150           There are platforms where longs are 64 bits, and platforms where
151           ints are 64 bits, and while we are out to shock you, even platforms
152           where shorts are 64 bits.  This is all legal according to the C
153           standard.  (In other words, "long long" is not a portable way to
154           specify 64 bits, and "long long" is not even guaranteed to be any
155           wider than "long".)
156
157           Instead, use the definitions IV, UV, IVSIZE, I32SIZE, and so forth.
158           Avoid things like I32 because they are not guaranteed to be exactly
159           32 bits, they are at least 32 bits, nor are they guaranteed to be
160           int or long.  If you really explicitly need 64-bit variables, use
161           I64 and U64, but only if guarded by HAS_QUAD.
162
163       •   Assuming one can dereference any type of pointer for any type of
164           data
165
166             char *p = ...;
167             long pony = *(long *)p;    /* BAD */
168
169           Many platforms, quite rightly so, will give you a core dump instead
170           of a pony if the p happens not to be correctly aligned.
171
172       •   Lvalue casts
173
174             (int)*p = ...;    /* BAD */
175
176           Simply not portable.  Get your lvalue to be of the right type, or
177           maybe use temporary variables, or dirty tricks with unions.
178
179       •   Assume anything about structs (especially the ones you don't
180           control, like the ones coming from the system headers)
181
182           •       That a certain field exists in a struct
183
184           •       That no other fields exist besides the ones you know of
185
186           •       That a field is of certain signedness, sizeof, or type
187
188           •       That the fields are in a certain order
189
190                   •       While C guarantees the ordering specified in the
191                           struct definition, between different platforms the
192                           definitions might differ
193
194           •       That the sizeof(struct) or the alignments are the same
195                   everywhere
196
197                   •       There might be padding bytes between the fields to
198                           align the fields - the bytes can be anything
199
200                   •       Structs are required to be aligned to the maximum
201                           alignment required by the fields - which for native
202                           types is for usually equivalent to sizeof() of the
203                           field
204
205       •   Assuming the character set is ASCIIish
206
207           Perl can compile and run under EBCDIC platforms.  See perlebcdic.
208           This is transparent for the most part, but because the character
209           sets differ, you shouldn't use numeric (decimal, octal, nor hex)
210           constants to refer to characters.  You can safely say 'A', but not
211           0x41.  You can safely say '\n', but not "\012".  However, you can
212           use macros defined in utf8.h to specify any code point portably.
213           "LATIN1_TO_NATIVE(0xDF)" is going to be the code point that means
214           LATIN SMALL LETTER SHARP S on whatever platform you are running on
215           (on ASCII platforms it compiles without adding any extra code, so
216           there is zero performance hit on those).  The acceptable inputs to
217           "LATIN1_TO_NATIVE" are from 0x00 through 0xFF.  If your input isn't
218           guaranteed to be in that range, use "UNICODE_TO_NATIVE" instead.
219           "NATIVE_TO_LATIN1" and "NATIVE_TO_UNICODE" translate the opposite
220           direction.
221
222           If you need the string representation of a character that doesn't
223           have a mnemonic name in C, you should add it to the list in
224           regen/unicode_constants.pl, and have Perl create "#define"'s for
225           you, based on the current platform.
226
227           Note that the "isFOO" and "toFOO" macros in handy.h work properly
228           on native code points and strings.
229
230           Also, the range 'A' - 'Z' in ASCII is an unbroken sequence of 26
231           upper case alphabetic characters.  That is not true in EBCDIC.  Nor
232           for 'a' to 'z'.  But '0' - '9' is an unbroken range in both
233           systems.  Don't assume anything about other ranges.  (Note that
234           special handling of ranges in regular expression patterns and
235           transliterations makes it appear to Perl code that the
236           aforementioned ranges are all unbroken.)
237
238           Many of the comments in the existing code ignore the possibility of
239           EBCDIC, and may be wrong therefore, even if the code works.  This
240           is actually a tribute to the successful transparent insertion of
241           being able to handle EBCDIC without having to change pre-existing
242           code.
243
244           UTF-8 and UTF-EBCDIC are two different encodings used to represent
245           Unicode code points as sequences of bytes.  Macros  with the same
246           names (but different definitions) in utf8.h and utfebcdic.h are
247           used to allow the calling code to think that there is only one such
248           encoding.  This is almost always referred to as "utf8", but it
249           means the EBCDIC version as well.  Again, comments in the code may
250           well be wrong even if the code itself is right.  For example, the
251           concept of UTF-8 "invariant characters" differs between ASCII and
252           EBCDIC.  On ASCII platforms, only characters that do not have the
253           high-order bit set (i.e.  whose ordinals are strict ASCII, 0 - 127)
254           are invariant, and the documentation and comments in the code may
255           assume that, often referring to something like, say, "hibit".  The
256           situation differs and is not so simple on EBCDIC machines, but as
257           long as the code itself uses the "NATIVE_IS_INVARIANT()" macro
258           appropriately, it works, even if the comments are wrong.
259
260           As noted in "TESTING" in perlhack, when writing test scripts, the
261           file t/charset_tools.pl contains some helpful functions for writing
262           tests valid on both ASCII and EBCDIC platforms.  Sometimes, though,
263           a test can't use a function and it's inconvenient to have different
264           test versions depending on the platform.  There are 20 code points
265           that are the same in all 4 character sets currently recognized by
266           Perl (the 3 EBCDIC code pages plus ISO 8859-1 (ASCII/Latin1)).
267           These can be used in such tests, though there is a small
268           possibility that Perl will become available in yet another
269           character set, breaking your test.  All but one of these code
270           points are C0 control characters.  The most significant controls
271           that are the same are "\0", "\r", and "\N{VT}" (also specifiable as
272           "\cK", "\x0B", "\N{U+0B}", or "\013").  The single non-control is
273           U+00B6 PILCROW SIGN.  The controls that are the same have the same
274           bit pattern in all 4 character sets, regardless of the UTF8ness of
275           the string containing them.  The bit pattern for U+B6 is the same
276           in all 4 for non-UTF8 strings, but differs in each when its
277           containing string is UTF-8 encoded.  The only other code points
278           that have some sort of sameness across all 4 character sets are the
279           pair 0xDC and 0xFC.  Together these represent upper- and lowercase
280           LATIN LETTER U WITH DIAERESIS, but which is upper and which is
281           lower may be reversed: 0xDC is the capital in Latin1 and 0xFC is
282           the small letter, while 0xFC is the capital in EBCDIC and 0xDC is
283           the small one.  This factoid may be exploited in writing case
284           insensitive tests that are the same across all 4 character sets.
285
286       •   Assuming the character set is just ASCII
287
288           ASCII is a 7 bit encoding, but bytes have 8 bits in them.  The 128
289           extra characters have different meanings depending on the locale.
290           Absent a locale, currently these extra characters are generally
291           considered to be unassigned, and this has presented some problems.
292           This has being changed starting in 5.12 so that these characters
293           can be considered to be Latin-1 (ISO-8859-1).
294
295       •   Mixing #define and #ifdef
296
297             #define BURGLE(x) ... \
298             #ifdef BURGLE_OLD_STYLE        /* BAD */
299             ... do it the old way ... \
300             #else
301             ... do it the new way ... \
302             #endif
303
304           You cannot portably "stack" cpp directives.  For example in the
305           above you need two separate BURGLE() #defines, one for each #ifdef
306           branch.
307
308       •   Adding non-comment stuff after #endif or #else
309
310             #ifdef SNOSH
311             ...
312             #else !SNOSH    /* BAD */
313             ...
314             #endif SNOSH    /* BAD */
315
316           The #endif and #else cannot portably have anything non-comment
317           after them.  If you want to document what is going (which is a good
318           idea especially if the branches are long), use (C) comments:
319
320             #ifdef SNOSH
321             ...
322             #else /* !SNOSH */
323             ...
324             #endif /* SNOSH */
325
326           The gcc option "-Wendif-labels" warns about the bad variant (by
327           default on starting from Perl 5.9.4).
328
329       •   Having a comma after the last element of an enum list
330
331             enum color {
332               CERULEAN,
333               CHARTREUSE,
334               CINNABAR,     /* BAD */
335             };
336
337           is not portable.  Leave out the last comma.
338
339           Also note that whether enums are implicitly morphable to ints
340           varies between compilers, you might need to (int).
341
342       •   Using //-comments
343
344             // This function bamfoodles the zorklator.   /* BAD */
345
346           That is C99 or C++.  Perl is C89.  Using the //-comments is
347           silently allowed by many C compilers but cranking up the ANSI C89
348           strictness (which we like to do) causes the compilation to fail.
349
350       •   Mixing declarations and code
351
352             void zorklator()
353             {
354               int n = 3;
355               set_zorkmids(n);    /* BAD */
356               int q = 4;
357
358           That is C99 or C++.  Some C compilers allow that, but you
359           shouldn't.
360
361           The gcc option "-Wdeclaration-after-statement" scans for such
362           problems (by default on starting from Perl 5.9.4).
363
364       •   Introducing variables inside for()
365
366             for(int i = ...; ...; ...) {    /* BAD */
367
368           That is C99 or C++.  While it would indeed be awfully nice to have
369           that also in C89, to limit the scope of the loop variable, alas, we
370           cannot.
371
372       •   Mixing signed char pointers with unsigned char pointers
373
374             int foo(char *s) { ... }
375             ...
376             unsigned char *t = ...; /* Or U8* t = ... */
377             foo(t);   /* BAD */
378
379           While this is legal practice, it is certainly dubious, and
380           downright fatal in at least one platform: for example VMS cc
381           considers this a fatal error.  One cause for people often making
382           this mistake is that a "naked char" and therefore dereferencing a
383           "naked char pointer" have an undefined signedness: it depends on
384           the compiler and the flags of the compiler and the underlying
385           platform whether the result is signed or unsigned.  For this very
386           same reason using a 'char' as an array index is bad.
387
388       •   Macros that have string constants and their arguments as substrings
389           of the string constants
390
391             #define FOO(n) printf("number = %d\n", n)    /* BAD */
392             FOO(10);
393
394           Pre-ANSI semantics for that was equivalent to
395
396             printf("10umber = %d\10");
397
398           which is probably not what you were expecting.  Unfortunately at
399           least one reasonably common and modern C compiler does "real
400           backward compatibility" here, in AIX that is what still happens
401           even though the rest of the AIX compiler is very happily C89.
402
403       •   Using printf formats for non-basic C types
404
405              IV i = ...;
406              printf("i = %d\n", i);    /* BAD */
407
408           While this might by accident work in some platform (where IV
409           happens to be an "int"), in general it cannot.  IV might be
410           something larger.  Even worse the situation is with more specific
411           types (defined by Perl's configuration step in config.h):
412
413              Uid_t who = ...;
414              printf("who = %d\n", who);    /* BAD */
415
416           The problem here is that Uid_t might be not only not "int"-wide but
417           it might also be unsigned, in which case large uids would be
418           printed as negative values.
419
420           There is no simple solution to this because of printf()'s limited
421           intelligence, but for many types the right format is available as
422           with either 'f' or '_f' suffix, for example:
423
424              IVdf /* IV in decimal */
425              UVxf /* UV is hexadecimal */
426
427              printf("i = %"IVdf"\n", i); /* The IVdf is a string constant. */
428
429              Uid_t_f /* Uid_t in decimal */
430
431              printf("who = %"Uid_t_f"\n", who);
432
433           Or you can try casting to a "wide enough" type:
434
435              printf("i = %"IVdf"\n", (IV)something_very_small_and_signed);
436
437           See "Formatted Printing of Size_t and SSize_t" in perlguts for how
438           to print those.
439
440           Also remember that the %p format really does require a void
441           pointer:
442
443              U8* p = ...;
444              printf("p = %p\n", (void*)p);
445
446           The gcc option "-Wformat" scans for such problems.
447
448       •   Blindly using variadic macros
449
450           gcc has had them for a while with its own syntax, and C99 brought
451           them with a standardized syntax.  Don't use the former, and use the
452           latter only if the HAS_C99_VARIADIC_MACROS is defined.
453
454       •   Blindly passing va_list
455
456           Not all platforms support passing va_list to further varargs
457           (stdarg) functions.  The right thing to do is to copy the va_list
458           using the Perl_va_copy() if the NEED_VA_COPY is defined.
459
460       •   Using gcc statement expressions
461
462              val = ({...;...;...});    /* BAD */
463
464           While a nice extension, it's not portable.  The Perl code does
465           admittedly use them if available to gain some extra speed
466           (essentially as a funky form of inlining), but you shouldn't.
467
468       •   Binding together several statements in a macro
469
470           Use the macros STMT_START and STMT_END.
471
472              STMT_START {
473                 ...
474              } STMT_END
475
476       •   Testing for operating systems or versions when should be testing
477           for features
478
479             #ifdef __FOONIX__    /* BAD */
480             foo = quux();
481             #endif
482
483           Unless you know with 100% certainty that quux() is only ever
484           available for the "Foonix" operating system and that is available
485           and correctly working for all past, present, and future versions of
486           "Foonix", the above is very wrong.  This is more correct (though
487           still not perfect, because the below is a compile-time check):
488
489             #ifdef HAS_QUUX
490             foo = quux();
491             #endif
492
493           How does the HAS_QUUX become defined where it needs to be?  Well,
494           if Foonix happens to be Unixy enough to be able to run the
495           Configure script, and Configure has been taught about detecting and
496           testing quux(), the HAS_QUUX will be correctly defined.  In other
497           platforms, the corresponding configuration step will hopefully do
498           the same.
499
500           In a pinch, if you cannot wait for Configure to be educated, or if
501           you have a good hunch of where quux() might be available, you can
502           temporarily try the following:
503
504             #if (defined(__FOONIX__) || defined(__BARNIX__))
505             # define HAS_QUUX
506             #endif
507
508             ...
509
510             #ifdef HAS_QUUX
511             foo = quux();
512             #endif
513
514           But in any case, try to keep the features and operating systems
515           separate.
516
517           A good resource on the predefined macros for various operating
518           systems, compilers, and so forth is
519           <http://sourceforge.net/p/predef/wiki/Home/>
520
521       •   Assuming the contents of static memory pointed to by the return
522           values of Perl wrappers for C library functions doesn't change.
523           Many C library functions return pointers to static storage that can
524           be overwritten by subsequent calls to the same or related
525           functions.  Perl has light-weight wrappers for some of these
526           functions, and which don't make copies of the static memory.  A
527           good example is the interface to the environment variables that are
528           in effect for the program.  Perl has "PerlEnv_getenv" to get values
529           from the environment.  But the return is a pointer to static memory
530           in the C library.  If you are using the value to immediately test
531           for something, that's fine, but if you save the value and expect it
532           to be unchanged by later processing, you would be wrong, but
533           perhaps you wouldn't know it because different C library
534           implementations behave differently, and the one on the platform
535           you're testing on might work for your situation.  But on some
536           platforms, a subsequent call to "PerlEnv_getenv" or related
537           function WILL overwrite the memory that your first call points to.
538           This has led to some hard-to-debug problems.  Do a "savepv" in
539           perlapi to make a copy, thus avoiding these problems.  You will
540           have to free the copy when you're done to avoid memory leaks.  If
541           you don't have control over when it gets freed, you'll need to make
542           the copy in a mortal scalar, like so:
543
544            if ((s = PerlEnv_getenv("foo") == NULL) {
545               ... /* handle NULL case */
546            }
547            else {
548                s = SvPVX(sv_2mortal(newSVpv(s, 0)));
549            }
550
551           The above example works only if "s" is "NUL"-terminated; otherwise
552           you have to pass its length to "newSVpv".
553
554   Problematic System Interfaces
555       •   Perl strings are NOT the same as C strings:  They may contain "NUL"
556           characters, whereas a C string is terminated by the first "NUL".
557           That is why Perl API functions that deal with strings generally
558           take a pointer to the first byte and either a length or a pointer
559           to the byte just beyond the final one.
560
561           And this is the reason that many of the C library string handling
562           functions should not be used.  They don't cope with the full
563           generality of Perl strings.  It may be that your test cases don't
564           have embedded "NUL"s, and so the tests pass, whereas there may well
565           eventually arise real-world cases where they fail.  A lesson here
566           is to include "NUL"s in your tests.  Now it's fairly rare in most
567           real world cases to get "NUL"s, so your code may seem to work,
568           until one day a "NUL" comes along.
569
570           Here's an example.  It used to be a common paradigm, for decades,
571           in the perl core to use "strchr("list", c)" to see if the character
572           "c" is any of the ones given in "list", a double-quote-enclosed
573           string of the set of characters that we are seeing if "c" is one
574           of.  As long as "c" isn't a "NUL", it works.  But when "c" is a
575           "NUL", "strchr" returns a pointer to the terminating "NUL" in
576           "list".   This likely will result in a segfault or a security issue
577           when the caller uses that end pointer as the starting point to read
578           from.
579
580           A solution to this and many similar issues is to use the "mem"-foo
581           C library functions instead.  In this case "memchr" can be used to
582           see if "c" is in "list" and works even if "c" is "NUL".  These
583           functions need an additional parameter to give the string length.
584           In the case of literal string parameters, perl has defined macros
585           that calculate the length for you.  See "String Handling" in
586           perlapi.
587
588       •   malloc(0), realloc(0), calloc(0, 0) are non-portable.  To be
589           portable allocate at least one byte.  (In general you should rarely
590           need to work at this low level, but instead use the various malloc
591           wrappers.)
592
593snprintf() - the return type is unportable.  Use my_snprintf()
594           instead.
595
596   Security problems
597       Last but not least, here are various tips for safer coding.  See also
598       perlclib for libc/stdio replacements one should use.
599
600       •   Do not use gets()
601
602           Or we will publicly ridicule you.  Seriously.
603
604       •   Do not use tmpfile()
605
606           Use mkstemp() instead.
607
608       •   Do not use strcpy() or strcat() or strncpy() or strncat()
609
610           Use my_strlcpy() and my_strlcat() instead: they either use the
611           native implementation, or Perl's own implementation (borrowed from
612           the public domain implementation of INN).
613
614       •   Do not use sprintf() or vsprintf()
615
616           If you really want just plain byte strings, use my_snprintf() and
617           my_vsnprintf() instead, which will try to use snprintf() and
618           vsnprintf() if those safer APIs are available.  If you want
619           something fancier than a plain byte string, use "Perl_form"() or
620           SVs and "Perl_sv_catpvf()".
621
622           Note that glibc "printf()", "sprintf()", etc. are buggy before
623           glibc version 2.17.  They won't allow a "%.s" format with a
624           precision to create a string that isn't valid UTF-8 if the current
625           underlying locale of the program is UTF-8.  What happens is that
626           the %s and its operand are simply skipped without any notice.
627           <https://sourceware.org/bugzilla/show_bug.cgi?id=6530>.
628
629       •   Do not use atoi()
630
631           Use grok_atoUV() instead.  atoi() has ill-defined behavior on
632           overflows, and cannot be used for incremental parsing.  It is also
633           affected by locale, which is bad.
634
635       •   Do not use strtol() or strtoul()
636
637           Use grok_atoUV() instead.  strtol() or strtoul() (or their
638           IV/UV-friendly macro disguises, Strtol() and Strtoul(), or Atol()
639           and Atoul() are affected by locale, which is bad.
640

DEBUGGING

642       You can compile a special debugging version of Perl, which allows you
643       to use the "-D" option of Perl to tell more about what Perl is doing.
644       But sometimes there is no alternative than to dive in with a debugger,
645       either to see the stack trace of a core dump (very useful in a bug
646       report), or trying to figure out what went wrong before the core dump
647       happened, or how did we end up having wrong or unexpected results.
648
649   Poking at Perl
650       To really poke around with Perl, you'll probably want to build Perl for
651       debugging, like this:
652
653           ./Configure -d -DDEBUGGING
654           make
655
656       "-DDEBUGGING" turns on the C compiler's "-g" flag to have it produce
657       debugging information which will allow us to step through a running
658       program, and to see in which C function we are at (without the
659       debugging information we might see only the numerical addresses of the
660       functions, which is not very helpful). It will also turn on the
661       "DEBUGGING" compilation symbol which enables all the internal debugging
662       code in Perl.  There are a whole bunch of things you can debug with
663       this: perlrun lists them all, and the best way to find out about them
664       is to play about with them.  The most useful options are probably
665
666           l  Context (loop) stack processing
667           s  Stack snapshots (with v, displays all stacks)
668           t  Trace execution
669           o  Method and overloading resolution
670           c  String/numeric conversions
671
672       For example
673
674           $ perl -Dst -e '$a + 1'
675           ....
676           (-e:1)      gvsv(main::a)
677               =>  UNDEF
678           (-e:1)      const(IV(1))
679               =>  UNDEF  IV(1)
680           (-e:1)      add
681               =>  NV(1)
682
683       Some of the functionality of the debugging code can be achieved with a
684       non-debugging perl by using XS modules:
685
686           -Dr => use re 'debug'
687           -Dx => use O 'Debug'
688
689   Using a source-level debugger
690       If the debugging output of "-D" doesn't help you, it's time to step
691       through perl's execution with a source-level debugger.
692
693       •  We'll use "gdb" for our examples here; the principles will apply to
694          any debugger (many vendors call their debugger "dbx"), but check the
695          manual of the one you're using.
696
697       To fire up the debugger, type
698
699           gdb ./perl
700
701       Or if you have a core dump:
702
703           gdb ./perl core
704
705       You'll want to do that in your Perl source tree so the debugger can
706       read the source code.  You should see the copyright message, followed
707       by the prompt.
708
709           (gdb)
710
711       "help" will get you into the documentation, but here are the most
712       useful commands:
713
714       •  run [args]
715
716          Run the program with the given arguments.
717
718       •  break function_name
719
720       •  break source.c:xxx
721
722          Tells the debugger that we'll want to pause execution when we reach
723          either the named function (but see "Internal Functions" in
724          perlguts!) or the given line in the named source file.
725
726       •  step
727
728          Steps through the program a line at a time.
729
730       •  next
731
732          Steps through the program a line at a time, without descending into
733          functions.
734
735       •  continue
736
737          Run until the next breakpoint.
738
739       •  finish
740
741          Run until the end of the current function, then stop again.
742
743       •  'enter'
744
745          Just pressing Enter will do the most recent operation again - it's a
746          blessing when stepping through miles of source code.
747
748       •  ptype
749
750          Prints the C definition of the argument given.
751
752            (gdb) ptype PL_op
753            type = struct op {
754                OP *op_next;
755                OP *op_sibparent;
756                OP *(*op_ppaddr)(void);
757                PADOFFSET op_targ;
758                unsigned int op_type : 9;
759                unsigned int op_opt : 1;
760                unsigned int op_slabbed : 1;
761                unsigned int op_savefree : 1;
762                unsigned int op_static : 1;
763                unsigned int op_folded : 1;
764                unsigned int op_spare : 2;
765                U8 op_flags;
766                U8 op_private;
767            } *
768
769       •  print
770
771          Execute the given C code and print its results.  WARNING: Perl makes
772          heavy use of macros, and gdb does not necessarily support macros
773          (see later "gdb macro support").  You'll have to substitute them
774          yourself, or to invoke cpp on the source code files (see "The .i
775          Targets") So, for instance, you can't say
776
777              print SvPV_nolen(sv)
778
779          but you have to say
780
781              print Perl_sv_2pv_nolen(sv)
782
783       You may find it helpful to have a "macro dictionary", which you can
784       produce by saying "cpp -dM perl.c | sort".  Even then, cpp won't
785       recursively apply those macros for you.
786
787   gdb macro support
788       Recent versions of gdb have fairly good macro support, but in order to
789       use it you'll need to compile perl with macro definitions included in
790       the debugging information.  Using gcc version 3.1, this means
791       configuring with "-Doptimize=-g3".  Other compilers might use a
792       different switch (if they support debugging macros at all).
793
794   Dumping Perl Data Structures
795       One way to get around this macro hell is to use the dumping functions
796       in dump.c; these work a little like an internal Devel::Peek, but they
797       also cover OPs and other structures that you can't get at from Perl.
798       Let's take an example.  We'll use the "$a = $b + $c" we used before,
799       but give it a bit of context: "$b = "6XXXX"; $c = 2.3;".  Where's a
800       good place to stop and poke around?
801
802       What about "pp_add", the function we examined earlier to implement the
803       "+" operator:
804
805           (gdb) break Perl_pp_add
806           Breakpoint 1 at 0x46249f: file pp_hot.c, line 309.
807
808       Notice we use "Perl_pp_add" and not "pp_add" - see "Internal Functions"
809       in perlguts.  With the breakpoint in place, we can run our program:
810
811           (gdb) run -e '$b = "6XXXX"; $c = 2.3; $a = $b + $c'
812
813       Lots of junk will go past as gdb reads in the relevant source files and
814       libraries, and then:
815
816           Breakpoint 1, Perl_pp_add () at pp_hot.c:309
817           1396    dSP; dATARGET; bool useleft; SV *svl, *svr;
818           (gdb) step
819           311           dPOPTOPnnrl_ul;
820           (gdb)
821
822       We looked at this bit of code before, and we said that "dPOPTOPnnrl_ul"
823       arranges for two "NV"s to be placed into "left" and "right" - let's
824       slightly expand it:
825
826        #define dPOPTOPnnrl_ul  NV right = POPn; \
827                                SV *leftsv = TOPs; \
828                                NV left = USE_LEFT(leftsv) ? SvNV(leftsv) : 0.0
829
830       "POPn" takes the SV from the top of the stack and obtains its NV either
831       directly (if "SvNOK" is set) or by calling the "sv_2nv" function.
832       "TOPs" takes the next SV from the top of the stack - yes, "POPn" uses
833       "TOPs" - but doesn't remove it.  We then use "SvNV" to get the NV from
834       "leftsv" in the same way as before - yes, "POPn" uses "SvNV".
835
836       Since we don't have an NV for $b, we'll have to use "sv_2nv" to convert
837       it.  If we step again, we'll find ourselves there:
838
839           (gdb) step
840           Perl_sv_2nv (sv=0xa0675d0) at sv.c:1669
841           1669        if (!sv)
842           (gdb)
843
844       We can now use "Perl_sv_dump" to investigate the SV:
845
846           (gdb) print Perl_sv_dump(sv)
847           SV = PV(0xa057cc0) at 0xa0675d0
848           REFCNT = 1
849           FLAGS = (POK,pPOK)
850           PV = 0xa06a510 "6XXXX"\0
851           CUR = 5
852           LEN = 6
853           $1 = void
854
855       We know we're going to get 6 from this, so let's finish the subroutine:
856
857           (gdb) finish
858           Run till exit from #0  Perl_sv_2nv (sv=0xa0675d0) at sv.c:1671
859           0x462669 in Perl_pp_add () at pp_hot.c:311
860           311           dPOPTOPnnrl_ul;
861
862       We can also dump out this op: the current op is always stored in
863       "PL_op", and we can dump it with "Perl_op_dump".  This'll give us
864       similar output to CPAN module B::Debug.
865
866           (gdb) print Perl_op_dump(PL_op)
867           {
868           13  TYPE = add  ===> 14
869               TARG = 1
870               FLAGS = (SCALAR,KIDS)
871               {
872                   TYPE = null  ===> (12)
873                     (was rv2sv)
874                   FLAGS = (SCALAR,KIDS)
875                   {
876           11          TYPE = gvsv  ===> 12
877                       FLAGS = (SCALAR)
878                       GV = main::b
879                   }
880               }
881
882       # finish this later #
883
884   Using gdb to look at specific parts of a program
885       With the example above, you knew to look for "Perl_pp_add", but what if
886       there were multiple calls to it all over the place, or you didn't know
887       what the op was you were looking for?
888
889       One way to do this is to inject a rare call somewhere near what you're
890       looking for.  For example, you could add "study" before your method:
891
892           study;
893
894       And in gdb do:
895
896           (gdb) break Perl_pp_study
897
898       And then step until you hit what you're looking for.  This works well
899       in a loop if you want to only break at certain iterations:
900
901           for my $c (1..100) {
902               study if $c == 50;
903           }
904
905   Using gdb to look at what the parser/lexer are doing
906       If you want to see what perl is doing when parsing/lexing your code,
907       you can use "BEGIN {}":
908
909           print "Before\n";
910           BEGIN { study; }
911           print "After\n";
912
913       And in gdb:
914
915           (gdb) break Perl_pp_study
916
917       If you want to see what the parser/lexer is doing inside of "if" blocks
918       and the like you need to be a little trickier:
919
920           if ($a && $b && do { BEGIN { study } 1 } && $c) { ... }
921

SOURCE CODE STATIC ANALYSIS

923       Various tools exist for analysing C source code statically, as opposed
924       to dynamically, that is, without executing the code.  It is possible to
925       detect resource leaks, undefined behaviour, type mismatches,
926       portability problems, code paths that would cause illegal memory
927       accesses, and other similar problems by just parsing the C code and
928       looking at the resulting graph, what does it tell about the execution
929       and data flows.  As a matter of fact, this is exactly how C compilers
930       know to give warnings about dubious code.
931
932   lint
933       The good old C code quality inspector, "lint", is available in several
934       platforms, but please be aware that there are several different
935       implementations of it by different vendors, which means that the flags
936       are not identical across different platforms.
937
938       There is a "lint" target in Makefile, but you may have to diddle with
939       the flags (see above).
940
941   Coverity
942       Coverity (<http://www.coverity.com/>) is a product similar to lint and
943       as a testbed for their product they periodically check several open
944       source projects, and they give out accounts to open source developers
945       to the defect databases.
946
947       There is Coverity setup for the perl5 project:
948       <https://scan.coverity.com/projects/perl5>
949
950   HP-UX cadvise (Code Advisor)
951       HP has a C/C++ static analyzer product for HP-UX caller Code Advisor.
952       (Link not given here because the URL is horribly long and seems
953       horribly unstable; use the search engine of your choice to find it.)
954       The use of the "cadvise_cc" recipe with "Configure ...
955       -Dcc=./cadvise_cc" (see cadvise "User Guide") is recommended; as is the
956       use of "+wall".
957
958   cpd (cut-and-paste detector)
959       The cpd tool detects cut-and-paste coding.  If one instance of the cut-
960       and-pasted code changes, all the other spots should probably be
961       changed, too.  Therefore such code should probably be turned into a
962       subroutine or a macro.
963
964       cpd (<http://pmd.sourceforge.net/cpd.html>) is part of the pmd project
965       (<http://pmd.sourceforge.net/>).  pmd was originally written for static
966       analysis of Java code, but later the cpd part of it was extended to
967       parse also C and C++.
968
969       Download the pmd-bin-X.Y.zip () from the SourceForge site, extract the
970       pmd-X.Y.jar from it, and then run that on source code thusly:
971
972         java -cp pmd-X.Y.jar net.sourceforge.pmd.cpd.CPD \
973          --minimum-tokens 100 --files /some/where/src --language c > cpd.txt
974
975       You may run into memory limits, in which case you should use the -Xmx
976       option:
977
978         java -Xmx512M ...
979
980   gcc warnings
981       Though much can be written about the inconsistency and coverage
982       problems of gcc warnings (like "-Wall" not meaning "all the warnings",
983       or some common portability problems not being covered by "-Wall", or
984       "-ansi" and "-pedantic" both being a poorly defined collection of
985       warnings, and so forth), gcc is still a useful tool in keeping our
986       coding nose clean.
987
988       The "-Wall" is by default on.
989
990       The "-ansi" (and its sidekick, "-pedantic") would be nice to be on
991       always, but unfortunately they are not safe on all platforms, they can
992       for example cause fatal conflicts with the system headers (Solaris
993       being a prime example).  If Configure "-Dgccansipedantic" is used, the
994       "cflags" frontend selects "-ansi -pedantic" for the platforms where
995       they are known to be safe.
996
997       The following extra flags are added:
998
999       •   "-Wendif-labels"
1000
1001       •   "-Wextra"
1002
1003       •   "-Wc++-compat"
1004
1005       •   "-Wwrite-strings"
1006
1007       •   "-Werror=declaration-after-statement"
1008
1009       •   "-Werror=pointer-arith"
1010
1011       The following flags would be nice to have but they would first need
1012       their own Augean stablemaster:
1013
1014       •   "-Wshadow"
1015
1016       •   "-Wstrict-prototypes"
1017
1018       The "-Wtraditional" is another example of the annoying tendency of gcc
1019       to bundle a lot of warnings under one switch (it would be impossible to
1020       deploy in practice because it would complain a lot) but it does contain
1021       some warnings that would be beneficial to have available on their own,
1022       such as the warning about string constants inside macros containing the
1023       macro arguments: this behaved differently pre-ANSI than it does in
1024       ANSI, and some C compilers are still in transition, AIX being an
1025       example.
1026
1027   Warnings of other C compilers
1028       Other C compilers (yes, there are other C compilers than gcc) often
1029       have their "strict ANSI" or "strict ANSI with some portability
1030       extensions" modes on, like for example the Sun Workshop has its "-Xa"
1031       mode on (though implicitly), or the DEC (these days, HP...) has its
1032       "-std1" mode on.
1033

MEMORY DEBUGGERS

1035       NOTE 1: Running under older memory debuggers such as Purify, valgrind
1036       or Third Degree greatly slows down the execution: seconds become
1037       minutes, minutes become hours.  For example as of Perl 5.8.1, the
1038       ext/Encode/t/Unicode.t takes extraordinarily long to complete under
1039       e.g. Purify, Third Degree, and valgrind.  Under valgrind it takes more
1040       than six hours, even on a snappy computer.  The said test must be doing
1041       something that is quite unfriendly for memory debuggers.  If you don't
1042       feel like waiting, that you can simply kill away the perl process.
1043       Roughly valgrind slows down execution by factor 10, AddressSanitizer by
1044       factor 2.
1045
1046       NOTE 2: To minimize the number of memory leak false alarms (see
1047       "PERL_DESTRUCT_LEVEL" for more information), you have to set the
1048       environment variable PERL_DESTRUCT_LEVEL to 2.  For example, like this:
1049
1050           env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib ...
1051
1052       NOTE 3: There are known memory leaks when there are compile-time errors
1053       within eval or require, seeing "S_doeval" in the call stack is a good
1054       sign of these.  Fixing these leaks is non-trivial, unfortunately, but
1055       they must be fixed eventually.
1056
1057       NOTE 4: DynaLoader will not clean up after itself completely unless
1058       Perl is built with the Configure option
1059       "-Accflags=-DDL_UNLOAD_ALL_AT_EXIT".
1060
1061   valgrind
1062       The valgrind tool can be used to find out both memory leaks and illegal
1063       heap memory accesses.  As of version 3.3.0, Valgrind only supports
1064       Linux on x86, x86-64 and PowerPC and Darwin (OS X) on x86 and x86-64.
1065       The special "test.valgrind" target can be used to run the tests under
1066       valgrind.  Found errors and memory leaks are logged in files named
1067       testfile.valgrind and by default output is displayed inline.
1068
1069       Example usage:
1070
1071           make test.valgrind
1072
1073       Since valgrind adds significant overhead, tests will take much longer
1074       to run.  The valgrind tests support being run in parallel to help with
1075       this:
1076
1077           TEST_JOBS=9 make test.valgrind
1078
1079       Note that the above two invocations will be very verbose as reachable
1080       memory and leak-checking is enabled by default.  If you want to just
1081       see pure errors, try:
1082
1083           VG_OPTS='-q --leak-check=no --show-reachable=no' TEST_JOBS=9 \
1084               make test.valgrind
1085
1086       Valgrind also provides a cachegrind tool, invoked on perl as:
1087
1088           VG_OPTS=--tool=cachegrind make test.valgrind
1089
1090       As system libraries (most notably glibc) are also triggering errors,
1091       valgrind allows to suppress such errors using suppression files.  The
1092       default suppression file that comes with valgrind already catches a lot
1093       of them.  Some additional suppressions are defined in t/perl.supp.
1094
1095       To get valgrind and for more information see
1096
1097           http://valgrind.org/
1098
1099   AddressSanitizer
1100       AddressSanitizer ("ASan") consists of a compiler instrumentation module
1101       and a run-time "malloc" library. ASan is available for a variety of
1102       architectures, operating systems, and compilers (see project link
1103       below).  It checks for unsafe memory usage, such as use after free and
1104       buffer overflow conditions, and is fast enough that you can easily
1105       compile your debugging or optimized perl with it. Modern versions of
1106       ASan check for memory leaks by default on most platforms, otherwise
1107       (e.g. x86_64 OS X) this feature can be enabled via
1108       "ASAN_OPTIONS=detect_leaks=1".
1109
1110       To build perl with AddressSanitizer, your Configure invocation should
1111       look like:
1112
1113           sh Configure -des -Dcc=clang \
1114              -Accflags=-fsanitize=address -Aldflags=-fsanitize=address \
1115              -Alddlflags=-shared\ -fsanitize=address \
1116              -fsanitize-blacklist=`pwd`/asan_ignore
1117
1118       where these arguments mean:
1119
1120       •   -Dcc=clang
1121
1122           This should be replaced by the full path to your clang executable
1123           if it is not in your path.
1124
1125       •   -Accflags=-fsanitize=address
1126
1127           Compile perl and extensions sources with AddressSanitizer.
1128
1129       •   -Aldflags=-fsanitize=address
1130
1131           Link the perl executable with AddressSanitizer.
1132
1133       •   -Alddlflags=-shared\ -fsanitize=address
1134
1135           Link dynamic extensions with AddressSanitizer.  You must manually
1136           specify "-shared" because using "-Alddlflags=-shared" will prevent
1137           Configure from setting a default value for "lddlflags", which
1138           usually contains "-shared" (at least on Linux).
1139
1140       •   -fsanitize-blacklist=`pwd`/asan_ignore
1141
1142           AddressSanitizer will ignore functions listed in the "asan_ignore"
1143           file. (This file should contain a short explanation of why each of
1144           the functions is listed.)
1145
1146       See also <https://github.com/google/sanitizers/wiki/AddressSanitizer>.
1147

PROFILING

1149       Depending on your platform there are various ways of profiling Perl.
1150
1151       There are two commonly used techniques of profiling executables:
1152       statistical time-sampling and basic-block counting.
1153
1154       The first method takes periodically samples of the CPU program counter,
1155       and since the program counter can be correlated with the code generated
1156       for functions, we get a statistical view of in which functions the
1157       program is spending its time.  The caveats are that very small/fast
1158       functions have lower probability of showing up in the profile, and that
1159       periodically interrupting the program (this is usually done rather
1160       frequently, in the scale of milliseconds) imposes an additional
1161       overhead that may skew the results.  The first problem can be
1162       alleviated by running the code for longer (in general this is a good
1163       idea for profiling), the second problem is usually kept in guard by the
1164       profiling tools themselves.
1165
1166       The second method divides up the generated code into basic blocks.
1167       Basic blocks are sections of code that are entered only in the
1168       beginning and exited only at the end.  For example, a conditional jump
1169       starts a basic block.  Basic block profiling usually works by
1170       instrumenting the code by adding enter basic block #nnnn book-keeping
1171       code to the generated code.  During the execution of the code the basic
1172       block counters are then updated appropriately.  The caveat is that the
1173       added extra code can skew the results: again, the profiling tools
1174       usually try to factor their own effects out of the results.
1175
1176   Gprof Profiling
1177       gprof is a profiling tool available in many Unix platforms which uses
1178       statistical time-sampling.  You can build a profiled version of perl by
1179       compiling using gcc with the flag "-pg".  Either edit config.sh or re-
1180       run Configure.  Running the profiled version of Perl will create an
1181       output file called gmon.out which contains the profiling data collected
1182       during the execution.
1183
1184       quick hint:
1185
1186           $ sh Configure -des -Dusedevel -Accflags='-pg' \
1187               -Aldflags='-pg' -Alddlflags='-pg -shared' \
1188               && make perl
1189           $ ./perl ... # creates gmon.out in current directory
1190           $ gprof ./perl > out
1191           $ less out
1192
1193       (you probably need to add "-shared" to the <-Alddlflags> line until RT
1194       #118199 is resolved)
1195
1196       The gprof tool can then display the collected data in various ways.
1197       Usually gprof understands the following options:
1198
1199       •   -a
1200
1201           Suppress statically defined functions from the profile.
1202
1203       •   -b
1204
1205           Suppress the verbose descriptions in the profile.
1206
1207       •   -e routine
1208
1209           Exclude the given routine and its descendants from the profile.
1210
1211       •   -f routine
1212
1213           Display only the given routine and its descendants in the profile.
1214
1215       •   -s
1216
1217           Generate a summary file called gmon.sum which then may be given to
1218           subsequent gprof runs to accumulate data over several runs.
1219
1220       •   -z
1221
1222           Display routines that have zero usage.
1223
1224       For more detailed explanation of the available commands and output
1225       formats, see your own local documentation of gprof.
1226
1227   GCC gcov Profiling
1228       basic block profiling is officially available in gcc 3.0 and later.
1229       You can build a profiled version of perl by compiling using gcc with
1230       the flags "-fprofile-arcs -ftest-coverage".  Either edit config.sh or
1231       re-run Configure.
1232
1233       quick hint:
1234
1235           $ sh Configure -des -Dusedevel -Doptimize='-g' \
1236               -Accflags='-fprofile-arcs -ftest-coverage' \
1237               -Aldflags='-fprofile-arcs -ftest-coverage' \
1238               -Alddlflags='-fprofile-arcs -ftest-coverage -shared' \
1239               && make perl
1240           $ rm -f regexec.c.gcov regexec.gcda
1241           $ ./perl ...
1242           $ gcov regexec.c
1243           $ less regexec.c.gcov
1244
1245       (you probably need to add "-shared" to the <-Alddlflags> line until RT
1246       #118199 is resolved)
1247
1248       Running the profiled version of Perl will cause profile output to be
1249       generated.  For each source file an accompanying .gcda file will be
1250       created.
1251
1252       To display the results you use the gcov utility (which should be
1253       installed if you have gcc 3.0 or newer installed).  gcov is run on
1254       source code files, like this
1255
1256           gcov sv.c
1257
1258       which will cause sv.c.gcov to be created.  The .gcov files contain the
1259       source code annotated with relative frequencies of execution indicated
1260       by "#" markers.  If you want to generate .gcov files for all profiled
1261       object files, you can run something like this:
1262
1263           for file in `find . -name \*.gcno`
1264           do sh -c "cd `dirname $file` && gcov `basename $file .gcno`"
1265           done
1266
1267       Useful options of gcov include "-b" which will summarise the basic
1268       block, branch, and function call coverage, and "-c" which instead of
1269       relative frequencies will use the actual counts.  For more information
1270       on the use of gcov and basic block profiling with gcc, see the latest
1271       GNU CC manual.  As of gcc 4.8, this is at
1272       <http://gcc.gnu.org/onlinedocs/gcc/Gcov-Intro.html#Gcov-Intro>
1273

MISCELLANEOUS TRICKS

1275   PERL_DESTRUCT_LEVEL
1276       If you want to run any of the tests yourself manually using e.g.
1277       valgrind, please note that by default perl does not explicitly cleanup
1278       all the memory it has allocated (such as global memory arenas) but
1279       instead lets the exit() of the whole program "take care" of such
1280       allocations, also known as "global destruction of objects".
1281
1282       There is a way to tell perl to do complete cleanup: set the environment
1283       variable PERL_DESTRUCT_LEVEL to a non-zero value.  The t/TEST wrapper
1284       does set this to 2, and this is what you need to do too, if you don't
1285       want to see the "global leaks": For example, for running under valgrind
1286
1287           env PERL_DESTRUCT_LEVEL=2 valgrind ./perl -Ilib t/foo/bar.t
1288
1289       (Note: the mod_perl apache module uses also this environment variable
1290       for its own purposes and extended its semantics.  Refer to the mod_perl
1291       documentation for more information.  Also, spawned threads do the
1292       equivalent of setting this variable to the value 1.)
1293
1294       If, at the end of a run you get the message N scalars leaked, you can
1295       recompile with "-DDEBUG_LEAKING_SCALARS", ("Configure
1296       -Accflags=-DDEBUG_LEAKING_SCALARS"), which will cause the addresses of
1297       all those leaked SVs to be dumped along with details as to where each
1298       SV was originally allocated.  This information is also displayed by
1299       Devel::Peek.  Note that the extra details recorded with each SV
1300       increases memory usage, so it shouldn't be used in production
1301       environments.  It also converts "new_SV()" from a macro into a real
1302       function, so you can use your favourite debugger to discover where
1303       those pesky SVs were allocated.
1304
1305       If you see that you're leaking memory at runtime, but neither valgrind
1306       nor "-DDEBUG_LEAKING_SCALARS" will find anything, you're probably
1307       leaking SVs that are still reachable and will be properly cleaned up
1308       during destruction of the interpreter.  In such cases, using the "-Dm"
1309       switch can point you to the source of the leak.  If the executable was
1310       built with "-DDEBUG_LEAKING_SCALARS", "-Dm" will output SV allocations
1311       in addition to memory allocations.  Each SV allocation has a distinct
1312       serial number that will be written on creation and destruction of the
1313       SV.  So if you're executing the leaking code in a loop, you need to
1314       look for SVs that are created, but never destroyed between each cycle.
1315       If such an SV is found, set a conditional breakpoint within "new_SV()"
1316       and make it break only when "PL_sv_serial" is equal to the serial
1317       number of the leaking SV.  Then you will catch the interpreter in
1318       exactly the state where the leaking SV is allocated, which is
1319       sufficient in many cases to find the source of the leak.
1320
1321       As "-Dm" is using the PerlIO layer for output, it will by itself
1322       allocate quite a bunch of SVs, which are hidden to avoid recursion.
1323       You can bypass the PerlIO layer if you use the SV logging provided by
1324       "-DPERL_MEM_LOG" instead.
1325
1326   PERL_MEM_LOG
1327       If compiled with "-DPERL_MEM_LOG" ("-Accflags=-DPERL_MEM_LOG"), both
1328       memory and SV allocations go through logging functions, which is handy
1329       for breakpoint setting.
1330
1331       Unless "-DPERL_MEM_LOG_NOIMPL" ("-Accflags=-DPERL_MEM_LOG_NOIMPL") is
1332       also compiled, the logging functions read $ENV{PERL_MEM_LOG} to
1333       determine whether to log the event, and if so how:
1334
1335           $ENV{PERL_MEM_LOG} =~ /m/           Log all memory ops
1336           $ENV{PERL_MEM_LOG} =~ /s/           Log all SV ops
1337           $ENV{PERL_MEM_LOG} =~ /t/           include timestamp in Log
1338           $ENV{PERL_MEM_LOG} =~ /^(\d+)/      write to FD given (default is 2)
1339
1340       Memory logging is somewhat similar to "-Dm" but is independent of
1341       "-DDEBUGGING", and at a higher level; all uses of Newx(), Renew(), and
1342       Safefree() are logged with the caller's source code file and line
1343       number (and C function name, if supported by the C compiler).  In
1344       contrast, "-Dm" is directly at the point of "malloc()".  SV logging is
1345       similar.
1346
1347       Since the logging doesn't use PerlIO, all SV allocations are logged and
1348       no extra SV allocations are introduced by enabling the logging.  If
1349       compiled with "-DDEBUG_LEAKING_SCALARS", the serial number for each SV
1350       allocation is also logged.
1351
1352   DDD over gdb
1353       Those debugging perl with the DDD frontend over gdb may find the
1354       following useful:
1355
1356       You can extend the data conversion shortcuts menu, so for example you
1357       can display an SV's IV value with one click, without doing any typing.
1358       To do that simply edit ~/.ddd/init file and add after:
1359
1360         ! Display shortcuts.
1361         Ddd*gdbDisplayShortcuts: \
1362         /t ()   // Convert to Bin\n\
1363         /d ()   // Convert to Dec\n\
1364         /x ()   // Convert to Hex\n\
1365         /o ()   // Convert to Oct(\n\
1366
1367       the following two lines:
1368
1369         ((XPV*) (())->sv_any )->xpv_pv  // 2pvx\n\
1370         ((XPVIV*) (())->sv_any )->xiv_iv // 2ivx
1371
1372       so now you can do ivx and pvx lookups or you can plug there the sv_peek
1373       "conversion":
1374
1375         Perl_sv_peek(my_perl, (SV*)()) // sv_peek
1376
1377       (The my_perl is for threaded builds.)  Just remember that every line,
1378       but the last one, should end with \n\
1379
1380       Alternatively edit the init file interactively via: 3rd mouse button ->
1381       New Display -> Edit Menu
1382
1383       Note: you can define up to 20 conversion shortcuts in the gdb section.
1384
1385   C backtrace
1386       On some platforms Perl supports retrieving the C level backtrace
1387       (similar to what symbolic debuggers like gdb do).
1388
1389       The backtrace returns the stack trace of the C call frames, with the
1390       symbol names (function names), the object names (like "perl"), and if
1391       it can, also the source code locations (file:line).
1392
1393       The supported platforms are Linux, and OS X (some *BSD might work at
1394       least partly, but they have not yet been tested).
1395
1396       This feature hasn't been tested with multiple threads, but it will only
1397       show the backtrace of the thread doing the backtracing.
1398
1399       The feature needs to be enabled with "Configure -Dusecbacktrace".
1400
1401       The "-Dusecbacktrace" also enables keeping the debug information when
1402       compiling/linking (often: "-g").  Many compilers/linkers do support
1403       having both optimization and keeping the debug information.  The debug
1404       information is needed for the symbol names and the source locations.
1405
1406       Static functions might not be visible for the backtrace.
1407
1408       Source code locations, even if available, can often be missing or
1409       misleading if the compiler has e.g. inlined code.  Optimizer can make
1410       matching the source code and the object code quite challenging.
1411
1412       Linux
1413           You must have the BFD (-lbfd) library installed, otherwise "perl"
1414           will fail to link.  The BFD is usually distributed as part of the
1415           GNU binutils.
1416
1417           Summary: "Configure ... -Dusecbacktrace" and you need "-lbfd".
1418
1419       OS X
1420           The source code locations are supported only if you have the
1421           Developer Tools installed.  (BFD is not needed.)
1422
1423           Summary: "Configure ... -Dusecbacktrace" and installing the
1424           Developer Tools would be good.
1425
1426       Optionally, for trying out the feature, you may want to enable
1427       automatic dumping of the backtrace just before a warning or croak (die)
1428       message is emitted, by adding "-Accflags=-DUSE_C_BACKTRACE_ON_ERROR"
1429       for Configure.
1430
1431       Unless the above additional feature is enabled, nothing about the
1432       backtrace functionality is visible, except for the Perl/XS level.
1433
1434       Furthermore, even if you have enabled this feature to be compiled, you
1435       need to enable it in runtime with an environment variable:
1436       "PERL_C_BACKTRACE_ON_ERROR=10".  It must be an integer higher than
1437       zero, telling the desired frame count.
1438
1439       Retrieving the backtrace from Perl level (using for example an XS
1440       extension) would be much less exciting than one would hope: normally
1441       you would see "runops", "entersub", and not much else.  This API is
1442       intended to be called from within the Perl implementation, not from
1443       Perl level execution.
1444
1445       The C API for the backtrace is as follows:
1446
1447       get_c_backtrace
1448       free_c_backtrace
1449       get_c_backtrace_dump
1450       dump_c_backtrace
1451
1452   Poison
1453       If you see in a debugger a memory area mysteriously full of 0xABABABAB
1454       or 0xEFEFEFEF, you may be seeing the effect of the Poison() macros, see
1455       perlclib.
1456
1457   Read-only optrees
1458       Under ithreads the optree is read only.  If you want to enforce this,
1459       to check for write accesses from buggy code, compile with
1460       "-Accflags=-DPERL_DEBUG_READONLY_OPS" to enable code that allocates op
1461       memory via "mmap", and sets it read-only when it is attached to a
1462       subroutine.  Any write access to an op results in a "SIGBUS" and abort.
1463
1464       This code is intended for development only, and may not be portable
1465       even to all Unix variants.  Also, it is an 80% solution, in that it
1466       isn't able to make all ops read only.  Specifically it does not apply
1467       to op slabs belonging to "BEGIN" blocks.
1468
1469       However, as an 80% solution it is still effective, as it has caught
1470       bugs in the past.
1471
1472   When is a bool not a bool?
1473       On pre-C99 compilers, "bool" is defined as equivalent to "char".
1474       Consequently assignment of any larger type to a "bool" is unsafe and
1475       may be truncated.  The "cBOOL" macro exists to cast it correctly; you
1476       may also find that using it is shorter and clearer than writing out the
1477       equivalent conditional expression longhand.
1478
1479       On those platforms and compilers where "bool" really is a boolean (C++,
1480       C99), it is easy to forget the cast.  You can force "bool" to be a
1481       "char" by compiling with "-Accflags=-DPERL_BOOL_AS_CHAR".  You may also
1482       wish to run "Configure" with something like
1483
1484           -Accflags='-Wconversion -Wno-sign-conversion -Wno-shorten-64-to-32'
1485
1486       or your compiler's equivalent to make it easier to spot any unsafe
1487       truncations that show up.
1488
1489       The "TRUE" and "FALSE" macros are available for situations where using
1490       them would clarify intent. (But they always just mean the same as the
1491       integers 1 and 0 regardless, so using them isn't compulsory.)
1492
1493   The .i Targets
1494       You can expand the macros in a foo.c file by saying
1495
1496           make foo.i
1497
1498       which will expand the macros using cpp.  Don't be scared by the
1499       results.
1500

AUTHOR

1502       This document was originally written by Nathan Torkington, and is
1503       maintained by the perl5-porters mailing list.
1504
1505
1506
1507perl v5.34.1                      2022-03-15                   PERLHACKTIPS(1)
Impressum