This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
[dummy merge]
[perl5.git] / Porting / pumpkin.pod
CommitLineData
aa689395 1=head1 NAME
2
3Pumpkin - Notes on handling the Perl Patch Pumpkin
4
5=head1 SYNOPSIS
6
7There is no simple synopsis, yet.
8
9=head1 DESCRIPTION
10
11This document attempts to begin to describe some of the
12considerations involved in patching and maintaining perl.
13
14This document is still under construction, and still subject to
15significant changes. Still, I hope parts of it will be useful,
16so I'm releasing it even though it's not done.
17
18For the most part, it's a collection of anecdotal information that
19already assumes some familiarity with the Perl sources. I really need
20an introductory section that describes the organization of the sources
21and all the various auxiliary files that are part of the distribution.
22
23=head1 Where Do I Get Perl Sources and Related Material?
24
25The Comprehensive Perl Archive Network (or CPAN) is the place to go.
26There are many mirrors, but the easiest thing to use is probably
27http://www.perl.com/CPAN, which automatically points you to a
28mirror site "close" to you.
29
30=head2 Perl5-porters mailing list
31
32The mailing list perl5-porters@perl.org
33is the main group working with the development of perl. If you're
34interested in all the latest developments, you should definitely
35subscribe. The list is high volume, but generally has a
36fairly low noise level.
37
38Subscribe by sending the message (in the body of your letter)
39
40 subscribe perl5-porters
41
42to perl5-porters-request@perl.org .
43
44=head1 How are Perl Releases Numbered?
45
46Perl version numbers are floating point numbers, such as 5.004. The
47major version number is 5, the minor version is '0', and '03' is the
48patchlevel. The version number is available as the magic variable $],
49and can be used in comparisons, e.g.
50
51 print "You've got an old perl\n" if $] < 5.002;
52
53(Observations about the imprecision of floating point numbers for
54representing reality probably have more relevance than you might
55imagine :-)
56
57You can also require particular version (or later) with
58
59 use 5.002;
60
61=head2 Subversions
62
63In addition, there may be "developer" sub-versions available. These
64are not official releases. They may contain unstable experimental
65features, and are subject to rapid change. Such developer
66sub-versions are numbered with sub-version numbers. For example,
67version 5.004_04 is the 4'th developer version built on top of
685.004. It might include the _01, _02, and _03 changes, but it
69also might not. Sub-versions are allowed to be subversive.
70
71These sub-versions can also be used as floating point numbers, so
72you can do things such as
73
74 print "You've got an unstable perl\n" if $] == 5.00403;
75
76You can also require particular version (or later) with
77
78 use 5.004_03; # the "_" is optional
79
80Sub-versions produced by the members of perl5-porters are usually
81available on CPAN in the F<src/5.0/unsupported> directory.
82
83=head2 Why such a complicated scheme?
84
85Two reasons, really. At least.
86
87First, we need some way to identify releases that are known to
88have new features that need testing and exploration. The
89subversion scheme does that nicely while fitting into the
90C<use 5.004;> mold.
91
92Second, since most of the folks who help maintain perl do so on a
93free-time voluntary basis, perl development does not proceed at a
94precise pace, though it always seems to be moving ahead quickly.
95We needed some way to pass around the "patch pumpkin" to allow
96different people chances to work on different aspects of the
97distribution without getting in each other's way. It wouldn't be
98constructive to have multiple people working on incompatible
99implementations of the same idea. Instead what was needed was
100some kind of "baton" or "token" to pass around so everyone knew
101whose turn was next.
102
103=head2 Why is it called the patch pumpkin?
104
105Chip Salzenberg gets credit for that, with a nod to his cow orker,
106David Croy. We had passed around various names (baton, token, hot
107potato) but none caught on. Then, Chip asked:
108
109[begin quote]
110
111 Who has the patch pumpkin?
112
113To explain: David Croy once told me once that at a previous job,
114there was one tape drive and multiple systems that used it for backups.
115But instead of some high-tech exclusion software, they used a low-tech
116method to prevent multiple simultaneous backups: a stuffed pumpkin.
117No one was allowed to make backups unless they had the "backup pumpkin".
118
119[end quote]
120
121The name has stuck.
122
123=head1 Philosophical Issues in Patching Perl
124
125There are no absolute rules, but there are some general guidelines I
126have tried to follow as I apply patches to the perl sources.
127(This section is still under construction.)
128
129=head2 Solve problems as generally as possible
130
131(I still have to think of a good example here.)
132
133=head2 Seek consensus on major changes
134
135If you are making big changes, don't do it in secret. Discuss the
136ideas in advance on perl5-porters.
137
138=head2 Keep the documentation up-to-date
139
140If your changes may affect how users use perl, then check to be sure
141that the documentation is in sync with your changes. Be sure to
142check all the files F<pod/*.pod> and also the F<INSTALL> document.
143
144Consider writing the appropriate documentation first and then
145implementing it to correspond to the documentation.
146
147=head2 Avoid machine-specific #ifdef's
148
149To the extent reasonable, try to avoid machine-specific #ifdef's in
150the sources. Instead, use feature-specific #ifdef's. The reason is
151that the machine-specific #ifdef's may not be valid across major
152releases of the operating system. Further, the feature-specific tests
153may help out folks on another platform who have the same problem.
154
155=head2 Allow for lots of testing
156
157We should never release a main version without testing it as a
158subversion first.
159
160=head2 Automate generatation of derivative files
161
162The F<embed.h>, F<keywords.h>, F<opcode.h>, and F<perltoc.pod> files
163are all automatically generated by perl scripts. In general, don't
164patch these directly; patch the data files instead.
165
166F<Configure> and F<config_h.SH> are also automatically generated by
167B<metaconfig>. In general, you should patch the metaconfig units
168instead of patching these files directly. However, minor changes to
169F<Configure> may be made in between major sync-ups with the metaconfig
170units, which tends to be complicated operations.
171
172=head1 How to Make a Distribution
173
174There really ought to be a 'make dist' target, but there isn't.
175The 'dist' suite of tools also contains a number of tools that I haven't
176learned how to use yet. Some of them may make this all a bit easier.
177
178Here are the steps I go through to prepare a patch & distribution.
179
180Lots of it could doubtless be automated but isn't.
181
182=head2 Announce your intentions
183
184First, you should volunteer out loud to take the patch pumpkin. It's
185generally counter-productive to have multiple people working in secret
186on the same thing.
187
188At the same time, announce what you plan to do with the patch pumpkin,
189to allow folks a chance to object or suggest alternatives, or do it for
190you. Naturally, the patch pumpkin holder ought to incorporate various
191bug fixes and documentation improvements that are posted while he or
192she has the pumpkin, but there might also be larger issues at stake.
193
194One of the precepts of the subversion idea is that we shouldn't give
195it to anyone unless we have some idea what you're going to do with
196it.
197
198=head2 refresh pod/perltoc.pod
199
200Presumably, you have done a full C<make> in your working source
201directory. Before you C<make spotless> (if you do), and if you have
202changed any documentation in any module or pod file, change to the
203F<pod> directory and run C<make toc>.
204
205=head2 update patchlevel.h
206
207Don't be shy about using the subversion number, even for a relatively
208modest patch. We've never even come close to using all 99 subversions,
209and it's better to have a distinctive number for your patch. If you
210need feedback on your patch, go ahead and issue it and promise to
211incorporate that feedback quickly (e.g. within 1 week) and send out a
212second patch.
213
214=head2 run metaconfig
215
216If you need to make changes to Configure or config_h.SH, it may be best to
217change the appropriate metaconfig units instead, and regenerate Configure.
218
219 metaconfig -m
220
221will regenerate Configure and config_h.SH. More information on
222obtaining and running metaconfig is in the F<U/README> file that comes
223with Perl's metaconfig units. Perl's metaconfig units should be
224available the same place you found this file. On CPAN, look under my
225directory F<id/ANDYD/> for a file such as F<5.003_07-02.U.tar.gz>.
226That file should be unpacked in your main perl source directory. It
227contains the files needed to run B<metaconfig> to reproduce Perl's
228Configure script.
229
230Alternatively, do consider if the F<*ish.h> files might be a better
231place for your changes.
232
233=head2 MANIFEST
234
235Make sure the MANIFEST is up-to-date. You can use dist's B<manicheck>
236program for this. You can also use
237
238 perl -MExtUtils::Manifest -e fullcheck
239
240to do half the job. This will make sure everything listed in MANIFEST
241is included in the distribution. dist's B<manicheck> command will
242also list extra files in the directory that are not listed in
243MANIFEST.
244
245The MANIFEST is normally sorted, with one exception. Perl includes
246both a F<Configure> script and a F<configure> script. The
247F<configure> script is a front-end to the main F<Configure>, but
248is there to aid folks who use autoconf-generated F<configure> files
249for other software. The problem is that F<Configure> and F<configure>
250are the same on case-insensitive file systems, so I deliberately put
251F<configure> first in the MANIFEST so that the extraction of
252F<Configure> will overwrite F<configure> and leave you with the
253correct script. (The F<configure> script must also have write
254permission for this to work, so it's the only file in the distribution
255I normally have with write permission.)
256
257If you are using metaconfig to regenerate Configure, then you should note
258that metaconfig actually uses MANIFEST.new, so you want to be sure
259MANIFEST.new is up-to-date too. I haven't found the MANIFEST/MANIFEST.new
260distinction particularly useful, but that's probably because I still haven't
261learned how to use the full suite of tools in the dist distribution.
262
263=head2 Check permissions
264
265All the tests in the t/ directory ought to be executable. The
266main makefile used to do a 'chmod t/*/*.t', but that resulted in
267a self-modifying distribution--something some users would strongly
268prefer to avoid. Probably, the F<t/TEST> script should check for this
269and do the chmod if needed, but it doesn't currently.
270
271In all, the following files should probably be executable:
272
273 Configure
274 configpm
275 configure
276 embed.pl
277 installperl
278 installman
279 keywords.pl
280 lib/splain
281 myconfig
282 opcode.pl
283 perly.fixer
284 t/TEST
285 t/*/*.t
286 *.SH
287 vms/ext/Stdio/test.pl
288 vms/ext/filespec.t
289 vms/fndvers.com
290 x2p/*.SH
291
292Other things ought to be readable, at least :-).
293
294Probably, the permissions for the files could be encoded in MANIFEST
295somehow, but I'm reluctant to change MANIFEST itself because that
296could break old scripts that use MANIFEST.
297
298I seem to recall that some SVR3 systems kept some sort of file that listed
299permissions for system files; something like that might be appropriate.
300
301=head2 Run Configure
302
303This will build a config.sh and config.h. You can skip this if you haven't
304changed Configure or config_h.SH at all.
305
306=head2 Update config_H
307
308The config_H file is provided to help those folks who can't run Configure.
309It is important to keep it up-to-date. If you have changed config_h.SH,
310those changes must be reflected in config_H as well. (The name config_H was
311chosen to distinguish the file from config.h even on case-insensitive file
312systems.) Simply edit the existing config_H file; keep the first few
313explanatory lines and then copy your new config.h below.
314
315It may also be necessary to update vms/config.vms and
316plan9/config.plan9, though you should be quite careful in doing so if
317you are not familiar with those systems. You might want to issue your
318patch with a promise to quickly issue a follow-up that handles those
319directories.
320
321=head2 make run_byacc
322
323If you have byacc-1.8.2 (available from CPAN), and if there have been
324changes to F<perly.y>, you can regenerate the F<perly.c> file. The
325run_byacc makefile target does this by running byacc and then applying
326some patches so that byacc dynamically allocates space, rather than
327having fixed limits. This patch is handled by the F<perly.fixer>
328script. Depending on the nature of the changes to F<perly.y>, you may
329or may not have to hand-edit the patch to apply correctly. If you do,
330you should include the edited patch in the new distribution. If you
331have byacc-1.9, the patch won't apply cleanly. Changes to the printf
332output statements mean the patch won't apply cleanly. Long ago I
333started to fix F<perly.fixer> to detect this, but I never completed the
334task.
335
336Some additional notes from Larry on this:
337
338Don't forget to regenerate perly.c.diff.
339
340 byacc perly.y
341 mv y.tab.c perly.c
342 patch perly.c <perly.c.diff
343 # manually apply any failed hunks
344 diff -c2 perly.c.orig perly.c >perly.c.diff
345
346One chunk of lines that often fails begins with
347
348 #line 29 "perly.y"
349
350and ends one line before
351
352 #define YYERRCODE 256
353
354This only happens when you add or remove a token type. I suppose this
355could be automated, but it doesn't happen very often nowadays.
356
357Larry
358
359=head2 make regen_headers
360
361The F<embed.h>, F<keywords.h>, and F<opcode.h> files are all automatically
362generated by perl scripts. Since the user isn't guaranteed to have a
363working perl, we can't require the user to generate them. Hence you have
364to, if you're making a distribution.
365
366I used to include rules like the following in the makefile:
367
368 # The following three header files are generated automatically
369 # The correct versions should be already supplied with the perl kit,
370 # in case you don't have perl or 'sh' available.
371 # The - is to ignore error return codes in case you have the source
372 # installed read-only or you don't have perl yet.
373 keywords.h: keywords.pl
374 @echo "Don't worry if this fails."
375 - perl keywords.pl
376
377
378However, I got lots of mail consisting of people worrying because the
379command failed. I eventually decided that I would save myself time
380and effort by manually running C<make regen_headers> myself rather
381than answering all the questions and complaints about the failing
382command.
383
384=head2 global.sym and interp.sym
385
386Make sure these files are up-to-date. Read the comments in these
387files and in perl_exp.SH to see what to do.
388
389=head2 Binary compatibility
390
391If you do change F<global.sym> or F<interp.sym>, think carefully about
392what you are doing. To the extent reasonable, we'd like to maintain
393souce and binary compatibility with older releases of perl. That way,
394extensions built under one version of perl will continue to work with
395new versions of perl.
396
397Of course, some incompatible changes may well be necessary. I'm just
398suggesting that we not make any such changes without thinking carefully
399about them first. If possible, we should provide
400backwards-compatibility stubs. There's a lot of XS code out there.
401Let's not force people to keep changing it.
402
403=head2 Changes
404
405Be sure to update the F<Changes> file. Try to include both an overall
406summary as well as detailed descriptions of the changes. Your
407audience will include bother developers and users, so describe
408user-visible changes (if any) in terms they will understand, not in
409code like "initialize foo variable in bar function".
410
411There are differing opinions on whether the detailed descriptions
412ought to go in the Changes file or whether they ought to be available
413separately in the patch file (or both). There is no disagreement that
414detailed descriptions ought to be easily available somewhere.
415
416=head2 OS/2-specific updates
417
418In the os2 directory is F<diff.configure>, a set of OS/2-specific
419diffs against B<Configure>. If you make changes to Configure, you may
420want to consider regenerating this diff file to save trouble for the
421OS/2 maintainer.
422
423=head2 VMS-specific updates
424
425If you have changed F<perly.y>, then you may want to update
426F<vms/perly_{h,c}.vms> by running C<perl vms/vms_yfix.pl>.
427
428The Perl version number appears in several places under F<vms>.
429It is courteous to update these versions. For example, if you are
430making 5.004_42, replace "5.00441" with "5.00442".
431
432=head2 Making the new distribution
433
434Suppose, for example, that you want to make version 5.004_08. Then you can
435do something like the following
436
437 mkdir ../perl5.004_08
438 awk '{print $1}' MANIFEST | cpio -pdm ../perl5.004_08
439 cd ../
440 tar cf perl5.004_08.tar perl5.004_08
441 gzip --best perl5.004_08.tar
442
443=head2 Making a new patch
444
445I find the F<makepatch> utility quite handy for making patches.
446You can obtain it from any CPAN archive under
447http://www.perl.com/CPAN/authors/Johan_Vromans/. The only
448difference between my version and the standard one is that I have mine
449do a
450
451 # Print a reassuring "End of Patch" note so people won't
452 # wonder if their mailer truncated patches.
453 print "\n\nEnd of Patch.\n";
454
455at the end. That's because I used to get questions from people asking if
456their mail was truncated.
457
458Here's how I generate a new patch. I'll use the hypothetical
4595.004_07 to 5.004_08 patch as an example.
460
461 # unpack perl5.004_07/
462 gzip -d -c perl5.004_07.tar.gz | tar -xof -
463 # unpack perl5.004_08/
464 gzip -d -c perl5.004_08.tar.gz | tar -xof -
465 makepatch perl5.004_07 perl5.004_08 > perl5.004_08.pat
466
467Makepatch will automatically generate appropriate B<rm> commands to remove
468deleted files. Unfortunately, it will not correctly set permissions
469for newly created files, so you may have to do so manually. For example,
470patch 5.003_04 created a new test F<t/op/gv.t> which needs to be executable,
471so at the top of the patch, I inserted the following lines:
472
473 # Make a new test
474 touch t/op/gv.t
475 chmod +x t/opt/gv.t
476
477Now, of course, my patch is now wrong because makepatch didn't know I
478was going to do that command, and it patched against /dev/null.
479
480So, what I do is sort out all such shell commands that need to be in the
481patch (including possible mv-ing of files, if needed) and put that in the
482shell commands at the top of the patch. Next, I delete all the patch parts
483of perl5.004_08.pat, leaving just the shell commands. Then, I do the
484following:
485
486 cd perl5.003_07
487 sh ../perl5.003_08.pat
488 cd ..
489 makepatch perl5.003_07 perl5.003_08 >> perl5.003_08.pat
490
491(Note the append to preserve my shell commands.)
492Now, my patch will line up with what the end users are going to do.
493
494=head2 Testing your patch
495
496It seems obvious, but be sure to test your patch. That is, verify that
497it produces exactly the same thing as your full distribution.
498
499 rm -rf perl5.003_07
500 gzip -d -c perl5.003_07.tar.gz | tar -xf -
501 cd perl5.003_07
502 sh ../perl5.003_08.pat
503 patch -p1 -N < ../perl5.003_08.pat
504 cd ..
505 gdiff -r perl5.003_07 perl5.003_08
506
507where B<gdiff> is GNU diff. Other diff's may also do recursive checking.
508
509=head2 More testing
510
511Again, it's obvious, but you should test your new version as widely as you
512can. You can be sure you'll hear about it quickly if your version doesn't
513work on both ANSI and pre-ANSI compilers, and on common systems such as
514SunOS 4.1.[34], Solaris, and Linux.
515
516If your changes include conditional code, try to test the different
517branches as thoroughly as you can. For example, if your system
518supports dynamic loading, you can also test static loading with
519
520 sh Configure -Uusedl
521
522You can also hand-tweak your config.h to try out different #ifdef
523branches.
524
525=head1 Common Gotcha's
526
527=over 4
528
529=item #elif
530
531The '#elif' preprocessor directive is not understood on all systems.
532Specifically, I know that Pyramids don't understand it. Thus instead of the
533simple
534
535 #if defined(I_FOO)
536 # include <foo.h>
537 #elif defined(I_BAR)
538 # include <bar.h>
539 #else
540 # include <fubar.h>
541 #endif
542
543You have to do the more Byzantine
544
545 #if defined(I_FOO)
546 # include <foo.h>
547 #else
548 # if defined(I_BAR)
549 # include <bar.h>
550 # else
551 # include <fubar.h>
552 # endif
553 #endif
554
555Incidentally, whitespace between the leading '#' and the preprocessor
556command is not guaranteed, but is very portable and you may use it freely.
557I think it makes things a bit more readable, especially once things get
558rather deeply nested. I also think that things should almost never get
559too deeply nested, so it ought to be a moot point :-)
560
561=item Probably Prefer POSIX
562
563It's often the case that you'll need to choose whether to do
564something the BSD-ish way or the POSIX-ish way. It's usually not
565a big problem when the two systems use different names for similar
566functions, such as memcmp() and bcmp(). The perl.h header file
567handles these by appropriate #defines, selecting the POSIX mem*()
568functions if available, but falling back on the b*() functions, if
569need be.
570
571More serious is the case where some brilliant person decided to
572use the same function name but give it a different meaning or
573calling sequence :-). getpgrp() and setpgrp() come to mind.
574These are a real problem on systems that aim for conformance to
575one standard (e.g. POSIX), but still try to support the other way
576of doing things (e.g. BSD). My general advice (still not really
577implemented in the source) is to do something like the following.
578Suppose there are two alternative versions, fooPOSIX() and
579fooBSD().
580
581 #ifdef HAS_FOOPOSIX
582 /* use fooPOSIX(); */
583 #else
584 # ifdef HAS_FOOBSD
585 /* try to emulate fooPOSIX() with fooBSD();
586 perhaps with the following: */
587 # define fooPOSIX fooBSD
588 # else
589 # /* Uh, oh. We have to supply our own. */
590 # define fooPOSIX Perl_fooPOSIX
591 # endif
592 #endif
593
594=item Think positively
595
596If you need to add an #ifdef test, it is usually easier to follow if you
597think positively, e.g.
598
599 #ifdef HAS_NEATO_FEATURE
600 /* use neato feature */
601 #else
602 /* use some fallback mechanism */
603 #endif
604
605rather than the more impenetrable
606
607 #ifndef MISSING_NEATO_FEATURE
608 /* Not missing it, so we must have it, so use it */
609 #else
610 /* Are missing it, so fall back on something else. */
611 #endif
612
613Of course for this toy example, there's not much difference. But when
614the #ifdef's start spanning a couple of screen fulls, and the #else's
615are marked something like
616
617 #else /* !MISSING_NEATO_FEATURE */
618
619I find it easy to get lost.
620
621=item Providing Missing Functions -- Problem
622
623Not all systems have all the neat functions you might want or need, so
624you might decide to be helpful and provide an emulation. This is
625sound in theory and very kind of you, but please be careful about what
626you name the function. Let me use the C<pause()> function as an
627illustration.
628
629Perl5.003 has the following in F<perl.h>
630
631 #ifndef HAS_PAUSE
632 #define pause() sleep((32767<<16)+32767)
633 #endif
634
635Configure sets HAS_PAUSE if the system has the pause() function, so
636this #define only kicks in if the pause() function is missing.
637Nice idea, right?
638
639Unfortunately, some systems apparently have a prototype for pause()
640in F<unistd.h>, but don't actually have the function in the library.
641(Or maybe they do have it in a library we're not using.)
642
643Thus, the compiler sees something like
644
645 extern int pause(void);
646 /* . . . */
647 #define pause() sleep((32767<<16)+32767)
648
649and dies with an error message. (Some compilers don't mind this;
650others apparently do.)
651
652To work around this, 5.003_03 and later have the following in perl.h:
653
654 /* Some unistd.h's give a prototype for pause() even though
655 HAS_PAUSE ends up undefined. This causes the #define
656 below to be rejected by the compiler. Sigh.
657 */
658 #ifdef HAS_PAUSE
659 # define Pause pause
660 #else
661 # define Pause() sleep((32767<<16)+32767)
662 #endif
663
664This works.
665
666The curious reader may wonder why I didn't do the following in
667F<util.c> instead:
668
669 #ifndef HAS_PAUSE
670 void pause()
671 {
672 sleep((32767<<16)+32767);
673 }
674 #endif
675
676That is, since the function is missing, just provide it.
677Then things would probably be been alright, it would seem.
678
679Well, almost. It could be made to work. The problem arises from the
680conflicting needs of dynamic loading and namespace protection.
681
682For dynamic loading to work on AIX (and VMS) we need to provide a list
683of symbols to be exported. This is done by the script F<perl_exp.SH>,
684which reads F<global.sym> and F<interp.sym>. Thus, the C<pause>
685symbol would have to be added to F<global.sym> So far, so good.
686
687On the other hand, one of the goals of Perl5 is to make it easy to
688either extend or embed perl and link it with other libraries. This
689means we have to be careful to keep the visible namespace "clean".
690That is, we don't want perl's global variables to conflict with
691those in the other application library. Although this work is still
692in progress, the way it is currently done is via the F<embed.h> file.
693This file is built from the F<global.sym> and F<interp.sym> files,
694since those files already list the globally visible symbols. If we
695had added C<pause> to global.sym, then F<embed.h> would contain the
696line
697
698 #define pause Perl_pause
699
700and calls to C<pause> in the perl sources would now point to
701C<Perl_pause>. Now, when B<ld> is run to build the F<perl> executable,
702it will go looking for C<perl_pause>, which probably won't exist in any
703of the standard libraries. Thus the build of perl will fail.
704
705Those systems where C<HAS_PAUSE> is not defined would be ok, however,
706since they would get a C<Perl_pause> function in util.c. The rest of
707the world would be in trouble.
708
709And yes, this scenario has happened. On SCO, the function C<chsize>
710is available. (I think it's in F<-lx>, the Xenix compatibility
711library.) Since the perl4 days (and possibly before), Perl has
712included a C<chsize> function that gets called something akin to
713
714 #ifndef HAS_CHSIZE
715 I32 chsize(fd, length)
716 /* . . . */
717 #endif
718
719When 5.003 added
720
721 #define chsize Perl_chsize
722
723to F<embed.h>, the compile started failing on SCO systems.
724
725The "fix" is to give the function a different name. The one
726implemented in 5.003_05 isn't optimal, but here's what was done:
727
728 #ifdef HAS_CHSIZE
729 # ifdef my_chsize /* Probably #defined to Perl_my_chsize in embed.h */
730 # undef my_chsize
731 # endif
732 # define my_chsize chsize
733 #endif
734
735My explanatory comment in patch 5.003_05 said:
736
737 Undef and then re-define my_chsize from Perl_my_chsize to
738 just plain chsize if this system HAS_CHSIZE. This probably only
739 applies to SCO. This shows the perils of having internal
740 functions with the same name as external library functions :-).
741
742Now, we can safely put C<my_chsize> in F<global.sym>, export it, and
743hide it with F<embed.h>.
744
745To be consistent with what I did for C<pause>, I probably should have
746called the new function C<Chsize>, rather than C<my_chsize>.
747However, the perl sources are quite inconsistent on this (Consider
748New, Mymalloc, and Myremalloc, to name just a few.)
749
750There is a problem with this fix, however, in that C<Perl_chsize>
751was available as a F<libperl.a> library function in 5.003, but it
752isn't available any more (as of 5.003_07). This means that we've
753broken binary compatibility. This is not good.
754
755=item Providing missing functions -- some ideas
756
757We currently don't have a standard way of handling such missing
758function names. Right now, I'm effectively thinking aloud about a
759solution. Some day, I'll try to formally propose a solution.
760
761Part of the problem is that we want to have some functions listed as
762exported but not have their names mangled by embed.h or possibly
763conflict with names in standard system headers. We actually already
764have such a list at the end of F<perl_exp.SH> (though that list is
765out-of-date):
766
767 # extra globals not included above.
768 cat <<END >> perl.exp
769 perl_init_ext
770 perl_init_fold
771 perl_init_i18nl14n
772 perl_alloc
773 perl_construct
774 perl_destruct
775 perl_free
776 perl_parse
777 perl_run
778 perl_get_sv
779 perl_get_av
780 perl_get_hv
781 perl_get_cv
782 perl_call_argv
783 perl_call_pv
784 perl_call_method
785 perl_call_sv
786 perl_requirepv
787 safecalloc
788 safemalloc
789 saferealloc
790 safefree
791
792This still needs much thought, but I'm inclined to think that one
793possible solution is to prefix all such functions with C<perl_> in the
794source and list them along with the other C<perl_*> functions in
795F<perl_exp.SH>.
796
797Thus, for C<chsize>, we'd do something like the following:
798
799 /* in perl.h */
800 #ifdef HAS_CHSIZE
801 # define perl_chsize chsize
802 #endif
803
804then in some file (e.g. F<util.c> or F<doio.c>) do
805
806 #ifndef HAS_CHSIZE
807 I32 perl_chsize(fd, length)
808 /* implement the function here . . . */
809 #endif
810
811Alternatively, we could just always use C<chsize> everywhere and move
812C<chsize> from F<global.sym> to the end of F<perl_exp.SH>. That would
813probably be fine as long as our C<chsize> function agreed with all the
814C<chsize> function prototypes in the various systems we'll be using.
815As long as the prototypes in actual use don't vary that much, this is
816probably a good alternative. (As a counter-example, note how Configure
817and perl have to go through hoops to find and use get Malloc_t and
818Free_t for C<malloc> and C<free>.)
819
820At the moment, this latter option is what I tend to prefer.
821
822=item All the world's a VAX
823
824Sorry, showing my age:-). Still, all the world is not BSD 4.[34],
825SVR4, or POSIX. Be aware that SVR3-derived systems are still quite
826common (do you have any idea how many systems run SCO?) If you don't
827have a bunch of v7 manuals handy, the metaconfig units (by default
828installed in F</usr/local/lib/dist/U>) are a good resource to look at
829for portability.
830
831=back
832
833=head1 Miscellaneous Topics
834
835=head2 Autoconf
836
837Why does perl use a metaconfig-generated Configure script instead of an
838autoconf-generated configure script?
839
840Metaconfig and autoconf are two tools with very similar purposes.
841Metaconfig is actually the older of the two, and was originally written
842by Larry Wall, while autoconf is probably now used in a wider variety of
843packages. The autoconf info file discusses the history of autoconf and
844how it came to be. The curious reader is referred there for further
845information.
846
847Overall, both tools are quite good, I think, and the choice of which one
848to use could be argued either way. In March, 1994, when I was just
849starting to work on Configure support for Perl5, I considered both
850autoconf and metaconfig, and eventually decided to use metaconfig for the
851following reasons:
852
853=over 4
854
855=item Compatibility with Perl4
856
857Perl4 used metaconfig, so many of the #ifdef's were already set up for
858metaconfig. Of course metaconfig had evolved some since Perl4's days,
859but not so much that it posed any serious problems.
860
861=item Metaconfig worked for me
862
863My system at the time was Interactive 2.2, a SVR3.2/386 derivative that
864also had some POSIX support. Metaconfig-generated Configure scripts
865worked fine for me on that system. On the other hand, autoconf-generated
866scripts usually didn't. (They did come quite close, though, in some
867cases.) At the time, I actually fetched a large number of GNU packages
868and checked. Not a single one configured and compiled correctly
869out-of-the-box with the system's cc compiler.
870
871=item Configure can be interactive
872
873With both autoconf and metaconfig, if the script works, everything is
874fine. However, one of my main problems with autoconf-generated scripts
875was that if it guessed wrong about something, it could be B<very> hard to
876go back and fix it. For example, autoconf always insisted on passing the
877-Xp flag to cc (to turn on POSIX behavior), even when that wasn't what I
878wanted or needed for that package. There was no way short of editing the
879configure script to turn this off. You couldn't just edit the resulting
880Makefile at the end because the -Xp flag influenced a number of other
881configure tests.
882
883Metaconfig's Configure scripts, on the other hand, can be interactive.
884Thus if Configure is guessing things incorrectly, you can go back and fix
885them. This isn't as important now as it was when we were actively
886developing Configure support for new features such as dynamic loading,
887but it's still useful occasionally.
888
889=item GPL
890
891At the time, autoconf-generated scripts were covered under the GNU Public
892License, and hence weren't suitable for inclusion with Perl, which has a
893different licensing policy. (Autoconf's licensing has since changed.)
894
895=item Modularity
896
897Metaconfig builds up Configure from a collection of discrete pieces
898called "units". You can override the standard behavior by supplying your
899own unit. With autoconf, you have to patch the standard files instead.
900I find the metaconfig "unit" method easier to work with. Others
901may find metaconfig's units clumsy to work with.
902
903=back
904
905=head2 @INC search order
906
907By default, the list of perl library directories in @INC is the
908following:
909
910 $archlib
911 $privlib
912 $sitearch
913 $sitelib
914
915Specifically, on my Solaris/x86 system, I run
916B<sh Configure -Dprefix=/opt/perl> and I have the following
917directories:
918
919 /opt/perl/lib/i86pc-solaris/5.00307
920 /opt/perl/lib
921 /opt/perl/lib/site_perl/i86pc-solaris
922 /opt/perl/lib/site_perl
923
924That is, perl's directories come first, followed by the site-specific
925directories.
926
927The site libraries come second to support the usage of extensions
928across perl versions. Read the relevant section in F<INSTALL> for
929more information. If we ever make $sitearch version-specific, this
930topic could be revisited.
931
932=head2 Why isn't there a directory to override Perl's library?
933
934Mainly because no one's gotten around to making one. Note that
935"making one" involves changing perl.c, Configure, config_h.SH (and
936associated files, see above), and I<documenting> it all in the
937INSTALL file.
938
939Apparently, most folks who want to override one of the standard library
940files simply do it by overwriting the standard library files.
941
942=head2 APPLLIB
943
944In the perl.c sources, you'll find an undocumented APPLLIB_EXP
945variable, sort of like PRIVLIB_EXP and ARCHLIB_EXP (which are
946documented in config_h.SH). Here's what APPLLIB_EXP is for, from
947a mail message from Larry:
948
949 The main intent of APPLLIB_EXP is for folks who want to send out a
950 version of Perl embedded in their product. They would set the symbol
951 to be the name of the library containing the files needed to run or to
952 support their particular application. This works at the "override"
953 level to make sure they get their own versions of any library code that
954 they absolutely must have configuration control over.
955
956 As such, I don't see any conflict with a sysadmin using it for a
957 override-ish sort of thing, when installing a generic Perl. It should
958 probably have been named something to do with overriding though. Since
959 it's undocumented we could still change it... :-)
960
961Given that it's already there, you can use it to override
962distribution modules. If you do
963
964 sh Configure -Dccflags='-DAPPLLIB_EXP=/my/override'
965
966then perl.c will put /my/override ahead of ARCHLIB and PRIVLIB.
967
968=head1 Upload Your Work to CPAN
969
970You can upload your work to CPAN if you have a CPAN id. Check out
971http://www.perl.com/CPAN/modules/04pause.html for information on
972_PAUSE_, the Perl Author's Upload Server.
973
974I typically upload both the patch file, e.g. F<perl5.004_08.pat.gz>
975and the full tar file, e.g. F<perl5.004_08.tar.gz>.
976
977If you want your patch to appear in the F<src/5.0/unsupported>
978directory on CPAN, send e-mail to the CPAN master librarian. (Check
979out http://www.perl.com/CPAN/CPAN.html).
980
981=head1 Help Save the World
982
983You should definitely announce your patch on the perl5-porters list.
984You should also consider announcing your patch on
985comp.lang.perl.announce, though you should make it quite clear that a
986subversion is not a production release, and be prepared to deal with
987people who will not read your disclaimer.
988
989=head1 Todo
990
991Here, in no particular order, are some Configure and build-related
992items that merit consideration. This list isn't exhaustive, it's just
993what I came up with off the top of my head.
994
995=head2 Good ideas waiting for round tuits
996
997=over 4
998
999=item installprefix
1000
1001I think we ought to support
1002
1003 Configure -Dinstallprefix=/blah/blah
1004
1005Currently, we support B<-Dprefix=/blah/blah>, but the changing the install
1006location has to be handled by something like the F<config.over> trick
1007described in F<INSTALL>. AFS users also are treated specially.
1008We should probably duplicate the metaconfig prefix stuff for an
1009install prefix.
1010
1011=item Configure -Dsrcdir=/blah/blah
1012
1013We should be able to emulate B<configure --srcdir>. Tom Tromey
1014tromey@creche.cygnus.com has submitted some patches to
1015the dist-users mailing list along these lines. Eventually, they ought
1016to get folded back into the main distribution.
1017
1018=item Hint file fixes
1019
1020Various hint files work around Configure problems. We ought to fix
1021Configure so that most of them aren't needed.
1022
1023=item Hint file information
1024
1025Some of the hint file information (particularly dynamic loading stuff)
1026ought to be fed back into the main metaconfig distribution.
1027
1028=back
1029
1030=head2 Probably good ideas waiting for round tuits
1031
1032=over 4
1033
1034=item GNU configure --options
1035
1036I've received sensible suggestions for --exec_prefix and other
1037GNU configure --options. It's not always obvious exactly what is
1038intended, but this merits investigation.
1039
1040=item make clean
1041
1042Currently, B<make clean> isn't all that useful, though
1043B<make realclean> and B<make distclean> are. This needs a bit of
1044thought and documentation before it gets cleaned up.
1045
1046=item Try gcc if cc fails
1047
1048Currently, we just give up.
1049
1050=item bypassing safe*alloc wrappers
1051
1052On some systems, it may be safe to call the system malloc directly
1053without going through the util.c safe* layers. (Such systems would
1054accept free(0), for example.) This might be a time-saver for systems
1055that already have a good malloc. (Recent Linux libc's apparently have
1056a nice malloc that is well-tuned for the system.)
1057
1058=back
1059
1060=head2 Vague possibilities
1061
1062=over 4
1063
1064=item Win95, WinNT, and Win32 support
1065
1066We need to get something into the distribution for 32-bit Windows.
1067I'm tired of all the private e-mail questions I get, and I'm saddened
1068that so many folks keep trying to reinvent the same wheel.
1069
1070=item MacPerl
1071
1072Get some of the Macintosh stuff folded back into the main
1073distribution.
1074
1075=item gconvert replacement
1076
1077Maybe include a replacement function that doesn't lose data in rare
1078cases of coercion between string and numerical values.
1079
1080=item long long
1081
1082Can we support C<long long> on systems where C<long long> is larger
1083than what we've been using for C<IV>? What if you can't C<sprintf>
1084a C<long long>?
1085
1086=item Improve makedepend
1087
1088The current makedepend process is clunky and annoyingly slow, but it
1089works for most folks. Alas, it assumes that there is a filename
1090$firstmakefile that the B<make> command will try to use before it uses
1091F<Makefile>. Such may not be the case for all B<make> commands,
1092particularly those on non-Unix systems.
1093
1094Probably some variant of the BSD F<.depend> file will be useful.
1095We ought to check how other packages do this, if they do it at all.
1096We could probably pre-generate the dependencies (with the exception of
1097malloc.o, which could probably be determined at F<Makefile.SH>
1098extraction time.
1099
1100=item GNU Makefile standard targets
1101
1102GNU software generally has standardized Makefile targets. Unless we
1103have good reason to do otherwise, I see no reason not to support them.
1104
1105=item File locking
1106
1107Somehow, straighten out, document, and implement lockf(), flock(),
1108and/or fcntl() file locking. It's a mess.
1109
1110=back
1111
1112=head1 AUTHOR
1113
1114Andy Dougherty <doughera@lafcol.lafayette.edu>.
1115
1116Additions by Chip Salzenberg <chip@atlantic.net>.
1117
1118All opinions expressed herein are those of the authorZ<>(s).
1119
1120=head1 LAST MODIFIED
1121
1122$Id: pumpkin.pod,v 1.8 1997/02/18 18:19:20 chip Released $