add stub docs for ext/B, other minor tweaks
[p5sagit/p5-mst-13.2.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
7b5757d1 27http://www.perl.com/CPAN/README.html , which automatically points you to a
aa689395 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
fb73857a 44Archives of the list are held at:
45
46 http://www.rosat.mpe-garching.mpg.de/mailing-lists/perl-porters/
47
aa689395 48=head1 How are Perl Releases Numbered?
49
7b5757d1 50Perl version numbers are floating point numbers, such as 5.004.
51(Observations about the imprecision of floating point numbers for
52representing reality probably have more relevance than you might
53imagine :-) The major version number is 5 and the '004' is the
54patchlevel. (Questions such as whether or not '004' is really a minor
55version number can safely be ignored.:)
56
57The version number is available as the magic variable $],
aa689395 58and can be used in comparisons, e.g.
59
60 print "You've got an old perl\n" if $] < 5.002;
61
aa689395 62You can also require particular version (or later) with
63
64 use 5.002;
65
7b5757d1 66At some point in the future, we may need to decide what to call the
67next big revision. In the .package file used by metaconfig to
68generate Configure, there are two variables that might be relevant:
69$baserev=5.0 and $package=perl5. At various times, I have suggested
70we might change them to $baserev=5.1 and $package=perl5.1 if want
71to signify a fairly major update. Or, we might want to jump to perl6.
72Let's worry about that problem when we get there.
73
aa689395 74=head2 Subversions
75
76In addition, there may be "developer" sub-versions available. These
77are not official releases. They may contain unstable experimental
78features, and are subject to rapid change. Such developer
79sub-versions are numbered with sub-version numbers. For example,
fb73857a 80version 5.003_04 is the 4'th developer version built on top of
815.003. It might include the _01, _02, and _03 changes, but it
82also might not. Sub-versions are allowed to be subversive. (But see
83the next section for recent changes.)
aa689395 84
85These sub-versions can also be used as floating point numbers, so
86you can do things such as
87
7b5757d1 88 print "You've got an unstable perl\n" if $] == 5.00303;
aa689395 89
90You can also require particular version (or later) with
91
7b5757d1 92 use 5.003_03; # the "_" is optional
aa689395 93
94Sub-versions produced by the members of perl5-porters are usually
95available on CPAN in the F<src/5.0/unsupported> directory.
96
7b5757d1 97=head2 Maintenance and Development Subversions
98
99As an experiment, starting with version 5.004, subversions _01 through
100_49 will be reserved for bug-fix maintenance releases, and subversions
101_50 through _99 will be available for unstable development versions.
102
103The separate bug-fix track is being established to allow us an easy
104way to distribute important bug fixes without waiting for the
105developers to untangle all the other problems in the current
106developer's release.
107
fb73857a 108Trial releases of bug-fix maintenance releases are announced on
109perl5-porters. Trial releases use the new subversion number (to avoid
110testers installing it over the previous release) and include a 'local
111patch' entry in patchlevel.h.
112
7b5757d1 113Watch for announcements of maintenance subversions in
114comp.lang.perl.announce.
115
aa689395 116=head2 Why such a complicated scheme?
117
118Two reasons, really. At least.
119
7b5757d1 120First, we need some way to identify and release collections of patches
121that are known to have new features that need testing and exploration. The
aa689395 122subversion scheme does that nicely while fitting into the
123C<use 5.004;> mold.
124
125Second, since most of the folks who help maintain perl do so on a
126free-time voluntary basis, perl development does not proceed at a
127precise pace, though it always seems to be moving ahead quickly.
128We needed some way to pass around the "patch pumpkin" to allow
129different people chances to work on different aspects of the
130distribution without getting in each other's way. It wouldn't be
131constructive to have multiple people working on incompatible
132implementations of the same idea. Instead what was needed was
133some kind of "baton" or "token" to pass around so everyone knew
134whose turn was next.
135
136=head2 Why is it called the patch pumpkin?
137
138Chip Salzenberg gets credit for that, with a nod to his cow orker,
139David Croy. We had passed around various names (baton, token, hot
140potato) but none caught on. Then, Chip asked:
141
142[begin quote]
143
144 Who has the patch pumpkin?
145
146To explain: David Croy once told me once that at a previous job,
147there was one tape drive and multiple systems that used it for backups.
148But instead of some high-tech exclusion software, they used a low-tech
149method to prevent multiple simultaneous backups: a stuffed pumpkin.
150No one was allowed to make backups unless they had the "backup pumpkin".
151
152[end quote]
153
154The name has stuck.
155
156=head1 Philosophical Issues in Patching Perl
157
158There are no absolute rules, but there are some general guidelines I
159have tried to follow as I apply patches to the perl sources.
160(This section is still under construction.)
161
162=head2 Solve problems as generally as possible
163
7b5757d1 164Never implement a specific restricted solution to a problem when you
165can solve the same problem in a more general, flexible way.
166
167For example, for dynamic loading to work on some SVR4 systems, we had
168to build a shared libperl.so library. In order to build "FAT" binaries
169on NeXT 4.0 systems, we had to build a special libperl library. Rather
170than continuing to build a contorted nest of special cases, I
171generalized the process of building libperl so that NeXT and SVR4 users
172could still get their work done, but others could build a shared
173libperl if they wanted to as well.
aa689395 174
175=head2 Seek consensus on major changes
176
177If you are making big changes, don't do it in secret. Discuss the
178ideas in advance on perl5-porters.
179
180=head2 Keep the documentation up-to-date
181
182If your changes may affect how users use perl, then check to be sure
183that the documentation is in sync with your changes. Be sure to
184check all the files F<pod/*.pod> and also the F<INSTALL> document.
185
186Consider writing the appropriate documentation first and then
7b5757d1 187implementing your change to correspond to the documentation.
aa689395 188
189=head2 Avoid machine-specific #ifdef's
190
191To the extent reasonable, try to avoid machine-specific #ifdef's in
192the sources. Instead, use feature-specific #ifdef's. The reason is
193that the machine-specific #ifdef's may not be valid across major
194releases of the operating system. Further, the feature-specific tests
195may help out folks on another platform who have the same problem.
196
197=head2 Allow for lots of testing
198
199We should never release a main version without testing it as a
200subversion first.
201
6877a1cf 202=head2 Test popular applications and modules.
203
204We should never release a main version without testing whether or not
205it breaks various popular modules and applications. A partial list of
206such things would include majordomo, metaconfig, apache, Tk, CGI,
207libnet, and libwww, to name just a few. Of course it's quite possible
208that some of those things will be just plain broken and need to be fixed,
209but, in general, we ought to try to avoid breaking widely-installed
210things.
211
7b5757d1 212=head2 Automate generation of derivative files
aa689395 213
214The F<embed.h>, F<keywords.h>, F<opcode.h>, and F<perltoc.pod> files
215are all automatically generated by perl scripts. In general, don't
216patch these directly; patch the data files instead.
217
218F<Configure> and F<config_h.SH> are also automatically generated by
219B<metaconfig>. In general, you should patch the metaconfig units
dfe9444c 220instead of patching these files directly. However, very minor changes to
aa689395 221F<Configure> may be made in between major sync-ups with the metaconfig
dfe9444c 222units, which tends to be complicated operations. But be careful, this
223can quickly spiral out of control. Running metaconfig is not really
224hard.
aa689395 225
a8119d38 226Finally, the sample files in the F<Porting/> subdirectory are
227generated automatically by the script F<U/mksample> included
228with the metaconfig units. See L<"run metaconfig"> below for
229information on obtaining the metaconfig units.
230
aa689395 231=head1 How to Make a Distribution
232
233There really ought to be a 'make dist' target, but there isn't.
234The 'dist' suite of tools also contains a number of tools that I haven't
235learned how to use yet. Some of them may make this all a bit easier.
236
237Here are the steps I go through to prepare a patch & distribution.
238
3e3baf6d 239Lots of it could doubtless be automated but isn't. The Porting/makerel
240(make release) perl script does now help automate some parts of it.
aa689395 241
242=head2 Announce your intentions
243
244First, you should volunteer out loud to take the patch pumpkin. It's
245generally counter-productive to have multiple people working in secret
246on the same thing.
247
248At the same time, announce what you plan to do with the patch pumpkin,
249to allow folks a chance to object or suggest alternatives, or do it for
250you. Naturally, the patch pumpkin holder ought to incorporate various
251bug fixes and documentation improvements that are posted while he or
252she has the pumpkin, but there might also be larger issues at stake.
253
254One of the precepts of the subversion idea is that we shouldn't give
7b5757d1 255the patch pumpkin to anyone unless we have some idea what he or she
256is going to do with it.
aa689395 257
258=head2 refresh pod/perltoc.pod
259
260Presumably, you have done a full C<make> in your working source
261directory. Before you C<make spotless> (if you do), and if you have
262changed any documentation in any module or pod file, change to the
263F<pod> directory and run C<make toc>.
264
3e3baf6d 265=head2 run installhtml to check the validity of the pod files
266
aa689395 267=head2 update patchlevel.h
268
269Don't be shy about using the subversion number, even for a relatively
270modest patch. We've never even come close to using all 99 subversions,
271and it's better to have a distinctive number for your patch. If you
272need feedback on your patch, go ahead and issue it and promise to
273incorporate that feedback quickly (e.g. within 1 week) and send out a
274second patch.
275
276=head2 run metaconfig
277
278If you need to make changes to Configure or config_h.SH, it may be best to
279change the appropriate metaconfig units instead, and regenerate Configure.
280
281 metaconfig -m
282
d562869c 283will regenerate Configure and config_h.SH. More information on obtaining
284and running metaconfig is in the F<U/README> file that comes with Perl's
285metaconfig units. Perl's metaconfig units should be available on CPAN.
286As of perl5.004_70, they were in the file F<mc_units-5.004_70-01.tar.gz>
287under http://www.perl.com/CPAN/authors/id/ANDYD/ . The mc_units tar
93341792 288file should be unpacked in your main perl source directory. (Note: those
289units were for 5.004_70. There have been significant changes since then.
d562869c 290Check for later versions or contact perl5-porters@perl.org to obtain a
291pointer to the current version.)
aa689395 292
293Alternatively, do consider if the F<*ish.h> files might be a better
294place for your changes.
295
296=head2 MANIFEST
297
298Make sure the MANIFEST is up-to-date. You can use dist's B<manicheck>
299program for this. You can also use
300
3e3baf6d 301 perl -w -MExtUtils::Manifest=fullcheck -e fullcheck
aa689395 302
3e3baf6d 303Both commands will also list extra files in the directory that are not
304listed in MANIFEST.
aa689395 305
bfb7748a 306The MANIFEST is normally sorted.
aa689395 307
308If you are using metaconfig to regenerate Configure, then you should note
309that metaconfig actually uses MANIFEST.new, so you want to be sure
310MANIFEST.new is up-to-date too. I haven't found the MANIFEST/MANIFEST.new
311distinction particularly useful, but that's probably because I still haven't
312learned how to use the full suite of tools in the dist distribution.
313
314=head2 Check permissions
315
316All the tests in the t/ directory ought to be executable. The
317main makefile used to do a 'chmod t/*/*.t', but that resulted in
318a self-modifying distribution--something some users would strongly
d562869c 319prefer to avoid. The F<t/TEST> script will check for this
320and do the chmod if needed, but the tests still ought to be
321executable.
aa689395 322
323In all, the following files should probably be executable:
324
325 Configure
326 configpm
32fcaa0b 327 configure.gnu
aa689395 328 embed.pl
329 installperl
330 installman
331 keywords.pl
aa689395 332 myconfig
333 opcode.pl
334 perly.fixer
335 t/TEST
336 t/*/*.t
337 *.SH
338 vms/ext/Stdio/test.pl
339 vms/ext/filespec.t
aa689395 340 x2p/*.SH
341
342Other things ought to be readable, at least :-).
343
344Probably, the permissions for the files could be encoded in MANIFEST
345somehow, but I'm reluctant to change MANIFEST itself because that
346could break old scripts that use MANIFEST.
347
348I seem to recall that some SVR3 systems kept some sort of file that listed
349permissions for system files; something like that might be appropriate.
350
351=head2 Run Configure
352
353This will build a config.sh and config.h. You can skip this if you haven't
693762b4 354changed Configure or config_h.SH at all. I use the following command
aa689395 355
693762b4 356 sh Configure -Dprefix=/opt/perl -Doptimize=-O -Dusethreads \
357 -Dcf_by='yourname' \
358 -Dcf_email='yourname@yourhost.yourplace.com' \
359 -Dperladmin='yourname@yourhost.yourplace.com' \
360 -Dmydomain='.yourplace.com' \
361 -Dmyhostname='yourhost' \
362 -des
aa689395 363
693762b4 364=head2 Update Porting/config.sh and Porting/config_H
dfe9444c 365
693762b4 366[XXX
367This section needs revision. We're currently working on easing
368the task of keeping the vms, win32, and plan9 config.sh info
369up-to-date. The plan is to use keep up-to-date 'canned' config.sh
370files in the appropriate subdirectories and then generate 'canned'
371config.h files for vms, win32, etc. from the generic config.sh file.
372This is to ease maintenance. When Configure gets updated, the parts
373sometimes get scrambled around, and the changes in config_H can
374sometimes be very hard to follow. config.sh, on the other hand, can
375safely be sorted, so it's easy to track (typically very small) changes
376to config.sh and then propoagate them to a canned 'config.h' by any
377number of means, including a perl script in win32/ or carrying
378config.sh and config_h.SH to a Unix system and running sh
379config_h.SH.)
380XXX]
381
382The Porting/config.sh and Porting/config_H files are provided to
383help those folks who can't run Configure. It is important to keep
384them up-to-date. If you have changed config_h.SH, those changes must
385be reflected in config_H as well. (The name config_H was chosen to
386distinguish the file from config.h even on case-insensitive file systems.)
387Simply edit the existing config_H file; keep the first few explanatory
388lines and then copy your new config.h below.
aa689395 389
d562869c 390It may also be necessary to update win32/config.?c, vms/config.vms and
aa689395 391plan9/config.plan9, though you should be quite careful in doing so if
392you are not familiar with those systems. You might want to issue your
393patch with a promise to quickly issue a follow-up that handles those
394directories.
395
396=head2 make run_byacc
397
398If you have byacc-1.8.2 (available from CPAN), and if there have been
399changes to F<perly.y>, you can regenerate the F<perly.c> file. The
400run_byacc makefile target does this by running byacc and then applying
401some patches so that byacc dynamically allocates space, rather than
402having fixed limits. This patch is handled by the F<perly.fixer>
403script. Depending on the nature of the changes to F<perly.y>, you may
404or may not have to hand-edit the patch to apply correctly. If you do,
405you should include the edited patch in the new distribution. If you
406have byacc-1.9, the patch won't apply cleanly. Changes to the printf
407output statements mean the patch won't apply cleanly. Long ago I
408started to fix F<perly.fixer> to detect this, but I never completed the
409task.
410
411Some additional notes from Larry on this:
412
e262e9be 413Don't forget to regenerate perly_c.diff.
aa689395 414
7b5757d1 415 byacc -d perly.y
aa689395 416 mv y.tab.c perly.c
e262e9be 417 patch perly.c <perly_c.diff
aa689395 418 # manually apply any failed hunks
e262e9be 419 diff -c2 perly.c.orig perly.c >perly_c.diff
aa689395 420
421One chunk of lines that often fails begins with
422
423 #line 29 "perly.y"
424
425and ends one line before
426
427 #define YYERRCODE 256
428
429This only happens when you add or remove a token type. I suppose this
430could be automated, but it doesn't happen very often nowadays.
431
432Larry
433
434=head2 make regen_headers
435
436The F<embed.h>, F<keywords.h>, and F<opcode.h> files are all automatically
437generated by perl scripts. Since the user isn't guaranteed to have a
438working perl, we can't require the user to generate them. Hence you have
439to, if you're making a distribution.
440
441I used to include rules like the following in the makefile:
442
443 # The following three header files are generated automatically
444 # The correct versions should be already supplied with the perl kit,
445 # in case you don't have perl or 'sh' available.
446 # The - is to ignore error return codes in case you have the source
447 # installed read-only or you don't have perl yet.
448 keywords.h: keywords.pl
449 @echo "Don't worry if this fails."
450 - perl keywords.pl
451
452
7b5757d1 453However, I got B<lots> of mail consisting of people worrying because the
aa689395 454command failed. I eventually decided that I would save myself time
455and effort by manually running C<make regen_headers> myself rather
456than answering all the questions and complaints about the failing
457command.
458
3e3baf6d 459=head2 global.sym, interp.sym and perlio.sym
aa689395 460
461Make sure these files are up-to-date. Read the comments in these
462files and in perl_exp.SH to see what to do.
463
464=head2 Binary compatibility
465
466If you do change F<global.sym> or F<interp.sym>, think carefully about
467what you are doing. To the extent reasonable, we'd like to maintain
468souce and binary compatibility with older releases of perl. That way,
469extensions built under one version of perl will continue to work with
470new versions of perl.
471
472Of course, some incompatible changes may well be necessary. I'm just
473suggesting that we not make any such changes without thinking carefully
474about them first. If possible, we should provide
475backwards-compatibility stubs. There's a lot of XS code out there.
476Let's not force people to keep changing it.
477
478=head2 Changes
479
480Be sure to update the F<Changes> file. Try to include both an overall
481summary as well as detailed descriptions of the changes. Your
3e3baf6d 482audience will include other developers and users, so describe
aa689395 483user-visible changes (if any) in terms they will understand, not in
484code like "initialize foo variable in bar function".
485
486There are differing opinions on whether the detailed descriptions
487ought to go in the Changes file or whether they ought to be available
488separately in the patch file (or both). There is no disagreement that
489detailed descriptions ought to be easily available somewhere.
490
2a26e2f1 491=head2 Todo
492
493The F<Todo> file contains a roughly-catgorized unordered list of
494aspects of Perl that could use enhancement, features that could be
495added, areas that could be cleaned up, and so on. During your term as
496pumpkin-holder, you will probably address some of these issues, and
497perhaps identify others which, while you decide not to address them
498this time around, may be tackled in the future. Update the file
499reflect the situation as it stands when you hand over the pumpkin.
500
501You might like, early in your pumpkin-holding career, to see if you
502can find champions for partiticular issues on the to-do list: an issue
503owned is an issue more likely to be resolved.
504
c4f23d77 505There are also some more porting-specific L<Todo> items later in this
506file.
507
aa689395 508=head2 OS/2-specific updates
509
510In the os2 directory is F<diff.configure>, a set of OS/2-specific
511diffs against B<Configure>. If you make changes to Configure, you may
512want to consider regenerating this diff file to save trouble for the
513OS/2 maintainer.
514
7b5757d1 515You can also consider the OS/2 diffs as reminders of portability
516things that need to be fixed in Configure.
517
aa689395 518=head2 VMS-specific updates
519
520If you have changed F<perly.y>, then you may want to update
521F<vms/perly_{h,c}.vms> by running C<perl vms/vms_yfix.pl>.
522
523The Perl version number appears in several places under F<vms>.
524It is courteous to update these versions. For example, if you are
525making 5.004_42, replace "5.00441" with "5.00442".
526
527=head2 Making the new distribution
528
529Suppose, for example, that you want to make version 5.004_08. Then you can
530do something like the following
531
532 mkdir ../perl5.004_08
533 awk '{print $1}' MANIFEST | cpio -pdm ../perl5.004_08
534 cd ../
535 tar cf perl5.004_08.tar perl5.004_08
536 gzip --best perl5.004_08.tar
537
3e3baf6d 538These steps, with extra checks, are automated by the Porting/makerel
539script.
540
aa689395 541=head2 Making a new patch
542
543I find the F<makepatch> utility quite handy for making patches.
544You can obtain it from any CPAN archive under
3e3baf6d 545http://www.perl.com/CPAN/authors/Johan_Vromans/ . There are a couple
546of differences between my version and the standard one. I have mine do
547a
aa689395 548
549 # Print a reassuring "End of Patch" note so people won't
550 # wonder if their mailer truncated patches.
551 print "\n\nEnd of Patch.\n";
552
3e3baf6d 553at the end. That's because I used to get questions from people asking
554if their mail was truncated.
555
556It also writes Index: lines which include the new directory prefix
557(change Index: print, approx line 294 or 310 depending on the version,
558to read: print PATCH ("Index: $newdir$new\n");). That helps patches
559work with more POSIX conformant patch programs.
aa689395 560
561Here's how I generate a new patch. I'll use the hypothetical
5625.004_07 to 5.004_08 patch as an example.
563
564 # unpack perl5.004_07/
565 gzip -d -c perl5.004_07.tar.gz | tar -xof -
566 # unpack perl5.004_08/
567 gzip -d -c perl5.004_08.tar.gz | tar -xof -
568 makepatch perl5.004_07 perl5.004_08 > perl5.004_08.pat
569
570Makepatch will automatically generate appropriate B<rm> commands to remove
571deleted files. Unfortunately, it will not correctly set permissions
572for newly created files, so you may have to do so manually. For example,
573patch 5.003_04 created a new test F<t/op/gv.t> which needs to be executable,
574so at the top of the patch, I inserted the following lines:
575
576 # Make a new test
577 touch t/op/gv.t
578 chmod +x t/opt/gv.t
579
580Now, of course, my patch is now wrong because makepatch didn't know I
581was going to do that command, and it patched against /dev/null.
582
583So, what I do is sort out all such shell commands that need to be in the
584patch (including possible mv-ing of files, if needed) and put that in the
585shell commands at the top of the patch. Next, I delete all the patch parts
586of perl5.004_08.pat, leaving just the shell commands. Then, I do the
587following:
588
7b5757d1 589 cd perl5.004_07
590 sh ../perl5.004_08.pat
aa689395 591 cd ..
7b5757d1 592 makepatch perl5.004_07 perl5.004_08 >> perl5.004_08.pat
aa689395 593
594(Note the append to preserve my shell commands.)
595Now, my patch will line up with what the end users are going to do.
596
597=head2 Testing your patch
598
599It seems obvious, but be sure to test your patch. That is, verify that
600it produces exactly the same thing as your full distribution.
601
7b5757d1 602 rm -rf perl5.004_07
603 gzip -d -c perl5.004_07.tar.gz | tar -xf -
604 cd perl5.004_07
605 sh ../perl5.004_08.pat
606 patch -p1 -N < ../perl5.004_08.pat
aa689395 607 cd ..
7b5757d1 608 gdiff -r perl5.004_07 perl5.004_08
aa689395 609
610where B<gdiff> is GNU diff. Other diff's may also do recursive checking.
611
612=head2 More testing
613
614Again, it's obvious, but you should test your new version as widely as you
615can. You can be sure you'll hear about it quickly if your version doesn't
616work on both ANSI and pre-ANSI compilers, and on common systems such as
617SunOS 4.1.[34], Solaris, and Linux.
618
619If your changes include conditional code, try to test the different
620branches as thoroughly as you can. For example, if your system
621supports dynamic loading, you can also test static loading with
622
623 sh Configure -Uusedl
624
625You can also hand-tweak your config.h to try out different #ifdef
626branches.
627
628=head1 Common Gotcha's
629
630=over 4
631
632=item #elif
633
634The '#elif' preprocessor directive is not understood on all systems.
635Specifically, I know that Pyramids don't understand it. Thus instead of the
636simple
637
638 #if defined(I_FOO)
639 # include <foo.h>
640 #elif defined(I_BAR)
641 # include <bar.h>
642 #else
643 # include <fubar.h>
644 #endif
645
646You have to do the more Byzantine
647
648 #if defined(I_FOO)
649 # include <foo.h>
650 #else
651 # if defined(I_BAR)
652 # include <bar.h>
653 # else
654 # include <fubar.h>
655 # endif
656 #endif
657
658Incidentally, whitespace between the leading '#' and the preprocessor
659command is not guaranteed, but is very portable and you may use it freely.
660I think it makes things a bit more readable, especially once things get
661rather deeply nested. I also think that things should almost never get
662too deeply nested, so it ought to be a moot point :-)
663
664=item Probably Prefer POSIX
665
666It's often the case that you'll need to choose whether to do
667something the BSD-ish way or the POSIX-ish way. It's usually not
668a big problem when the two systems use different names for similar
669functions, such as memcmp() and bcmp(). The perl.h header file
670handles these by appropriate #defines, selecting the POSIX mem*()
671functions if available, but falling back on the b*() functions, if
672need be.
673
674More serious is the case where some brilliant person decided to
675use the same function name but give it a different meaning or
676calling sequence :-). getpgrp() and setpgrp() come to mind.
677These are a real problem on systems that aim for conformance to
678one standard (e.g. POSIX), but still try to support the other way
679of doing things (e.g. BSD). My general advice (still not really
680implemented in the source) is to do something like the following.
681Suppose there are two alternative versions, fooPOSIX() and
682fooBSD().
683
684 #ifdef HAS_FOOPOSIX
685 /* use fooPOSIX(); */
686 #else
687 # ifdef HAS_FOOBSD
688 /* try to emulate fooPOSIX() with fooBSD();
689 perhaps with the following: */
690 # define fooPOSIX fooBSD
691 # else
692 # /* Uh, oh. We have to supply our own. */
693 # define fooPOSIX Perl_fooPOSIX
694 # endif
695 #endif
696
697=item Think positively
698
699If you need to add an #ifdef test, it is usually easier to follow if you
700think positively, e.g.
701
702 #ifdef HAS_NEATO_FEATURE
703 /* use neato feature */
704 #else
705 /* use some fallback mechanism */
706 #endif
707
708rather than the more impenetrable
709
710 #ifndef MISSING_NEATO_FEATURE
711 /* Not missing it, so we must have it, so use it */
712 #else
713 /* Are missing it, so fall back on something else. */
714 #endif
715
716Of course for this toy example, there's not much difference. But when
717the #ifdef's start spanning a couple of screen fulls, and the #else's
718are marked something like
719
720 #else /* !MISSING_NEATO_FEATURE */
721
722I find it easy to get lost.
723
724=item Providing Missing Functions -- Problem
725
726Not all systems have all the neat functions you might want or need, so
727you might decide to be helpful and provide an emulation. This is
728sound in theory and very kind of you, but please be careful about what
729you name the function. Let me use the C<pause()> function as an
730illustration.
731
732Perl5.003 has the following in F<perl.h>
733
734 #ifndef HAS_PAUSE
735 #define pause() sleep((32767<<16)+32767)
736 #endif
737
738Configure sets HAS_PAUSE if the system has the pause() function, so
739this #define only kicks in if the pause() function is missing.
740Nice idea, right?
741
742Unfortunately, some systems apparently have a prototype for pause()
743in F<unistd.h>, but don't actually have the function in the library.
744(Or maybe they do have it in a library we're not using.)
745
746Thus, the compiler sees something like
747
748 extern int pause(void);
749 /* . . . */
750 #define pause() sleep((32767<<16)+32767)
751
752and dies with an error message. (Some compilers don't mind this;
753others apparently do.)
754
755To work around this, 5.003_03 and later have the following in perl.h:
756
757 /* Some unistd.h's give a prototype for pause() even though
758 HAS_PAUSE ends up undefined. This causes the #define
759 below to be rejected by the compiler. Sigh.
760 */
761 #ifdef HAS_PAUSE
762 # define Pause pause
763 #else
764 # define Pause() sleep((32767<<16)+32767)
765 #endif
766
767This works.
768
769The curious reader may wonder why I didn't do the following in
770F<util.c> instead:
771
772 #ifndef HAS_PAUSE
773 void pause()
774 {
775 sleep((32767<<16)+32767);
776 }
777 #endif
778
779That is, since the function is missing, just provide it.
780Then things would probably be been alright, it would seem.
781
782Well, almost. It could be made to work. The problem arises from the
783conflicting needs of dynamic loading and namespace protection.
784
785For dynamic loading to work on AIX (and VMS) we need to provide a list
786of symbols to be exported. This is done by the script F<perl_exp.SH>,
787which reads F<global.sym> and F<interp.sym>. Thus, the C<pause>
788symbol would have to be added to F<global.sym> So far, so good.
789
790On the other hand, one of the goals of Perl5 is to make it easy to
791either extend or embed perl and link it with other libraries. This
792means we have to be careful to keep the visible namespace "clean".
793That is, we don't want perl's global variables to conflict with
794those in the other application library. Although this work is still
795in progress, the way it is currently done is via the F<embed.h> file.
796This file is built from the F<global.sym> and F<interp.sym> files,
797since those files already list the globally visible symbols. If we
798had added C<pause> to global.sym, then F<embed.h> would contain the
799line
800
801 #define pause Perl_pause
802
803and calls to C<pause> in the perl sources would now point to
804C<Perl_pause>. Now, when B<ld> is run to build the F<perl> executable,
805it will go looking for C<perl_pause>, which probably won't exist in any
806of the standard libraries. Thus the build of perl will fail.
807
808Those systems where C<HAS_PAUSE> is not defined would be ok, however,
809since they would get a C<Perl_pause> function in util.c. The rest of
810the world would be in trouble.
811
812And yes, this scenario has happened. On SCO, the function C<chsize>
813is available. (I think it's in F<-lx>, the Xenix compatibility
814library.) Since the perl4 days (and possibly before), Perl has
815included a C<chsize> function that gets called something akin to
816
817 #ifndef HAS_CHSIZE
818 I32 chsize(fd, length)
819 /* . . . */
820 #endif
821
822When 5.003 added
823
824 #define chsize Perl_chsize
825
826to F<embed.h>, the compile started failing on SCO systems.
827
828The "fix" is to give the function a different name. The one
829implemented in 5.003_05 isn't optimal, but here's what was done:
830
831 #ifdef HAS_CHSIZE
832 # ifdef my_chsize /* Probably #defined to Perl_my_chsize in embed.h */
833 # undef my_chsize
834 # endif
835 # define my_chsize chsize
836 #endif
837
838My explanatory comment in patch 5.003_05 said:
839
840 Undef and then re-define my_chsize from Perl_my_chsize to
841 just plain chsize if this system HAS_CHSIZE. This probably only
842 applies to SCO. This shows the perils of having internal
843 functions with the same name as external library functions :-).
844
845Now, we can safely put C<my_chsize> in F<global.sym>, export it, and
846hide it with F<embed.h>.
847
848To be consistent with what I did for C<pause>, I probably should have
849called the new function C<Chsize>, rather than C<my_chsize>.
850However, the perl sources are quite inconsistent on this (Consider
851New, Mymalloc, and Myremalloc, to name just a few.)
852
853There is a problem with this fix, however, in that C<Perl_chsize>
854was available as a F<libperl.a> library function in 5.003, but it
855isn't available any more (as of 5.003_07). This means that we've
856broken binary compatibility. This is not good.
857
858=item Providing missing functions -- some ideas
859
860We currently don't have a standard way of handling such missing
861function names. Right now, I'm effectively thinking aloud about a
862solution. Some day, I'll try to formally propose a solution.
863
864Part of the problem is that we want to have some functions listed as
865exported but not have their names mangled by embed.h or possibly
866conflict with names in standard system headers. We actually already
867have such a list at the end of F<perl_exp.SH> (though that list is
868out-of-date):
869
870 # extra globals not included above.
871 cat <<END >> perl.exp
872 perl_init_ext
873 perl_init_fold
874 perl_init_i18nl14n
875 perl_alloc
876 perl_construct
877 perl_destruct
878 perl_free
879 perl_parse
880 perl_run
881 perl_get_sv
882 perl_get_av
883 perl_get_hv
884 perl_get_cv
885 perl_call_argv
886 perl_call_pv
887 perl_call_method
888 perl_call_sv
889 perl_requirepv
890 safecalloc
891 safemalloc
892 saferealloc
893 safefree
894
895This still needs much thought, but I'm inclined to think that one
896possible solution is to prefix all such functions with C<perl_> in the
897source and list them along with the other C<perl_*> functions in
898F<perl_exp.SH>.
899
900Thus, for C<chsize>, we'd do something like the following:
901
902 /* in perl.h */
903 #ifdef HAS_CHSIZE
904 # define perl_chsize chsize
905 #endif
906
907then in some file (e.g. F<util.c> or F<doio.c>) do
908
909 #ifndef HAS_CHSIZE
910 I32 perl_chsize(fd, length)
911 /* implement the function here . . . */
912 #endif
913
914Alternatively, we could just always use C<chsize> everywhere and move
915C<chsize> from F<global.sym> to the end of F<perl_exp.SH>. That would
916probably be fine as long as our C<chsize> function agreed with all the
917C<chsize> function prototypes in the various systems we'll be using.
918As long as the prototypes in actual use don't vary that much, this is
919probably a good alternative. (As a counter-example, note how Configure
920and perl have to go through hoops to find and use get Malloc_t and
921Free_t for C<malloc> and C<free>.)
922
923At the moment, this latter option is what I tend to prefer.
924
925=item All the world's a VAX
926
927Sorry, showing my age:-). Still, all the world is not BSD 4.[34],
928SVR4, or POSIX. Be aware that SVR3-derived systems are still quite
929common (do you have any idea how many systems run SCO?) If you don't
930have a bunch of v7 manuals handy, the metaconfig units (by default
931installed in F</usr/local/lib/dist/U>) are a good resource to look at
932for portability.
933
934=back
935
936=head1 Miscellaneous Topics
937
938=head2 Autoconf
939
940Why does perl use a metaconfig-generated Configure script instead of an
941autoconf-generated configure script?
942
943Metaconfig and autoconf are two tools with very similar purposes.
944Metaconfig is actually the older of the two, and was originally written
945by Larry Wall, while autoconf is probably now used in a wider variety of
946packages. The autoconf info file discusses the history of autoconf and
947how it came to be. The curious reader is referred there for further
948information.
949
950Overall, both tools are quite good, I think, and the choice of which one
951to use could be argued either way. In March, 1994, when I was just
952starting to work on Configure support for Perl5, I considered both
953autoconf and metaconfig, and eventually decided to use metaconfig for the
954following reasons:
955
956=over 4
957
958=item Compatibility with Perl4
959
960Perl4 used metaconfig, so many of the #ifdef's were already set up for
961metaconfig. Of course metaconfig had evolved some since Perl4's days,
962but not so much that it posed any serious problems.
963
964=item Metaconfig worked for me
965
966My system at the time was Interactive 2.2, a SVR3.2/386 derivative that
967also had some POSIX support. Metaconfig-generated Configure scripts
968worked fine for me on that system. On the other hand, autoconf-generated
969scripts usually didn't. (They did come quite close, though, in some
970cases.) At the time, I actually fetched a large number of GNU packages
971and checked. Not a single one configured and compiled correctly
972out-of-the-box with the system's cc compiler.
973
974=item Configure can be interactive
975
976With both autoconf and metaconfig, if the script works, everything is
977fine. However, one of my main problems with autoconf-generated scripts
978was that if it guessed wrong about something, it could be B<very> hard to
979go back and fix it. For example, autoconf always insisted on passing the
980-Xp flag to cc (to turn on POSIX behavior), even when that wasn't what I
981wanted or needed for that package. There was no way short of editing the
982configure script to turn this off. You couldn't just edit the resulting
983Makefile at the end because the -Xp flag influenced a number of other
984configure tests.
985
986Metaconfig's Configure scripts, on the other hand, can be interactive.
987Thus if Configure is guessing things incorrectly, you can go back and fix
988them. This isn't as important now as it was when we were actively
989developing Configure support for new features such as dynamic loading,
990but it's still useful occasionally.
991
992=item GPL
993
994At the time, autoconf-generated scripts were covered under the GNU Public
995License, and hence weren't suitable for inclusion with Perl, which has a
996different licensing policy. (Autoconf's licensing has since changed.)
997
998=item Modularity
999
1000Metaconfig builds up Configure from a collection of discrete pieces
1001called "units". You can override the standard behavior by supplying your
1002own unit. With autoconf, you have to patch the standard files instead.
1003I find the metaconfig "unit" method easier to work with. Others
1004may find metaconfig's units clumsy to work with.
1005
1006=back
1007
1008=head2 @INC search order
1009
1010By default, the list of perl library directories in @INC is the
1011following:
1012
1013 $archlib
1014 $privlib
1015 $sitearch
1016 $sitelib
1017
1018Specifically, on my Solaris/x86 system, I run
1019B<sh Configure -Dprefix=/opt/perl> and I have the following
1020directories:
1021
1022 /opt/perl/lib/i86pc-solaris/5.00307
1023 /opt/perl/lib
1024 /opt/perl/lib/site_perl/i86pc-solaris
1025 /opt/perl/lib/site_perl
1026
1027That is, perl's directories come first, followed by the site-specific
1028directories.
1029
1030The site libraries come second to support the usage of extensions
1031across perl versions. Read the relevant section in F<INSTALL> for
1032more information. If we ever make $sitearch version-specific, this
1033topic could be revisited.
1034
1035=head2 Why isn't there a directory to override Perl's library?
1036
1037Mainly because no one's gotten around to making one. Note that
1038"making one" involves changing perl.c, Configure, config_h.SH (and
1039associated files, see above), and I<documenting> it all in the
1040INSTALL file.
1041
1042Apparently, most folks who want to override one of the standard library
1043files simply do it by overwriting the standard library files.
1044
1045=head2 APPLLIB
1046
1047In the perl.c sources, you'll find an undocumented APPLLIB_EXP
1048variable, sort of like PRIVLIB_EXP and ARCHLIB_EXP (which are
1049documented in config_h.SH). Here's what APPLLIB_EXP is for, from
1050a mail message from Larry:
1051
1052 The main intent of APPLLIB_EXP is for folks who want to send out a
1053 version of Perl embedded in their product. They would set the symbol
1054 to be the name of the library containing the files needed to run or to
1055 support their particular application. This works at the "override"
1056 level to make sure they get their own versions of any library code that
1057 they absolutely must have configuration control over.
1058
1059 As such, I don't see any conflict with a sysadmin using it for a
1060 override-ish sort of thing, when installing a generic Perl. It should
1061 probably have been named something to do with overriding though. Since
1062 it's undocumented we could still change it... :-)
1063
1064Given that it's already there, you can use it to override
1065distribution modules. If you do
1066
1067 sh Configure -Dccflags='-DAPPLLIB_EXP=/my/override'
1068
1069then perl.c will put /my/override ahead of ARCHLIB and PRIVLIB.
1070
c4f23d77 1071=head2 Shared libperl.so location
1072
1073Why isn't the shared libperl.so installed in /usr/lib/ along
1074with "all the other" shared libraries? Instead, it is installed
1075in $archlib, which is typically something like
1076
1077 /usr/local/lib/perl5/archname/5.00404
1078
1079and is architecture- and version-specific.
1080
1081The basic reason why a shared libperl.so gets put in $archlib is so that
1082you can have more than one version of perl on the system at the same time,
1083and have each refer to its own libperl.so.
1084
1085Three examples might help. All of these work now; none would work if you
1086put libperl.so in /usr/lib.
1087
1088=over
1089
1090=item 1.
1091
1092Suppose you want to have both threaded and non-threaded perl versions
1093around. Configure will name both perl libraries "libperl.so" (so that
1094you can link to them with -lperl). The perl binaries tell them apart
1095by having looking in the appropriate $archlib directories.
1096
1097=item 2.
1098
1099Suppose you have perl5.004_04 installed and you want to try to compile
1100it again, perhaps with different options or after applying a patch.
1101If you already have libperl.so installed in /usr/lib/, then it may be
1102either difficult or impossible to get ld.so to find the new libperl.so
1103that you're trying to build. If, instead, libperl.so is tucked away in
1104$archlib, then you can always just change $archlib in the current perl
1105you're trying to build so that ld.so won't find your old libperl.so.
1106(The INSTALL file suggests you do this when building a debugging perl.)
1107
1108=item 3.
1109
1110The shared perl library is not a "well-behaved" shared library with
1111proper major and minor version numbers, so you can't necessarily
1112have perl5.004_04 and perl5.004_05 installed simultaneously. Suppose
1113perl5.004_04 were to install /usr/lib/libperl.so.4.4, and perl5.004_05
1114were to install /usr/lib/libperl.so.4.5. Now, when you try to run
1115perl5.004_04, ld.so might try to load libperl.so.4.5, since it has
1116the right "major version" number. If this works at all, it almost
1117certainly defeats the reason for keeping perl5.004_04 around. Worse,
1118with development subversions, you certaily can't guarantee that
1119libperl.so.4.4 and libperl.so.4.55 will be compatible.
1120
1121Anyway, all this leads to quite obscure failures that are sure to drive
1122casual users crazy. Even experienced users will get confused :-). Upon
1123reflection, I'd say leave libperl.so in $archlib.
1124
1125=back
1126
aa689395 1127=head1 Upload Your Work to CPAN
1128
1129You can upload your work to CPAN if you have a CPAN id. Check out
1130http://www.perl.com/CPAN/modules/04pause.html for information on
1131_PAUSE_, the Perl Author's Upload Server.
1132
1133I typically upload both the patch file, e.g. F<perl5.004_08.pat.gz>
1134and the full tar file, e.g. F<perl5.004_08.tar.gz>.
1135
1136If you want your patch to appear in the F<src/5.0/unsupported>
1137directory on CPAN, send e-mail to the CPAN master librarian. (Check
7b5757d1 1138out http://www.perl.com/CPAN/CPAN.html ).
aa689395 1139
1140=head1 Help Save the World
1141
1142You should definitely announce your patch on the perl5-porters list.
1143You should also consider announcing your patch on
1144comp.lang.perl.announce, though you should make it quite clear that a
1145subversion is not a production release, and be prepared to deal with
1146people who will not read your disclaimer.
1147
1148=head1 Todo
1149
1150Here, in no particular order, are some Configure and build-related
1151items that merit consideration. This list isn't exhaustive, it's just
1152what I came up with off the top of my head.
1153
1154=head2 Good ideas waiting for round tuits
1155
1156=over 4
1157
1158=item installprefix
1159
1160I think we ought to support
1161
1162 Configure -Dinstallprefix=/blah/blah
1163
1164Currently, we support B<-Dprefix=/blah/blah>, but the changing the install
1165location has to be handled by something like the F<config.over> trick
1166described in F<INSTALL>. AFS users also are treated specially.
1167We should probably duplicate the metaconfig prefix stuff for an
1168install prefix.
1169
c4f23d77 1170=item Configure -Dsrc=/blah/blah
aa689395 1171
1172We should be able to emulate B<configure --srcdir>. Tom Tromey
1173tromey@creche.cygnus.com has submitted some patches to
c4f23d77 1174the dist-users mailing list along these lines. They have been folded
1175back into the main distribution, but various parts of the perl
1176Configure/build/install process still assume src='.'.
aa689395 1177
1178=item Hint file fixes
1179
1180Various hint files work around Configure problems. We ought to fix
1181Configure so that most of them aren't needed.
1182
1183=item Hint file information
1184
1185Some of the hint file information (particularly dynamic loading stuff)
1186ought to be fed back into the main metaconfig distribution.
1187
c4f23d77 1188=item Catch GNU Libc "Stub" functions
1189
1190Some functions (such as lchown()) are present in libc, but are
1191unimplmented. That is, they always fail and set errno=ENOSYS.
1192
1193Thomas Bushnell provided the following sample code and the explanation
1194that follows:
1195
1196 /* System header to define __stub macros and hopefully few prototypes,
1197 which can conflict with char FOO(); below. */
1198 #include <assert.h>
1199 /* Override any gcc2 internal prototype to avoid an error. */
1200 /* We use char because int might match the return type of a gcc2
1201 builtin and then its argument prototype would still apply. */
1202 char FOO();
1203
1204 int main() {
1205
1206 /* The GNU C library defines this for functions which it implements
1207 to always fail with ENOSYS. Some functions are actually named
1208 something starting with __ and the normal name is an alias. */
1209 #if defined (__stub_FOO) || defined (__stub___FOO)
1210 choke me
1211 #else
1212 FOO();
1213 #endif
1214
1215 ; return 0; }
1216
1217The choice of <assert.h> is essentially arbitrary. The GNU libc
1218macros are found in <gnu/stubs.h>. You can include that file instead
1219of <assert.h> (which itself includes <gnu/stubs.h>) if you test for
1220its existence first. <assert.h> is assumed to exist on every system,
1221which is why it's used here. Any GNU libc header file will include
1222the stubs macros. If either __stub_NAME or __stub___NAME is defined,
1223then the function doesn't actually exist. Tests using <assert.h> work
1224on every system around.
1225
1226The declaration of FOO is there to override builtin prototypes for
1227ANSI C functions.
1228
aa689395 1229=back
1230
1231=head2 Probably good ideas waiting for round tuits
1232
1233=over 4
1234
1235=item GNU configure --options
1236
1237I've received sensible suggestions for --exec_prefix and other
1238GNU configure --options. It's not always obvious exactly what is
1239intended, but this merits investigation.
1240
1241=item make clean
1242
1243Currently, B<make clean> isn't all that useful, though
1244B<make realclean> and B<make distclean> are. This needs a bit of
1245thought and documentation before it gets cleaned up.
1246
1247=item Try gcc if cc fails
1248
1249Currently, we just give up.
1250
1251=item bypassing safe*alloc wrappers
1252
1253On some systems, it may be safe to call the system malloc directly
1254without going through the util.c safe* layers. (Such systems would
1255accept free(0), for example.) This might be a time-saver for systems
1256that already have a good malloc. (Recent Linux libc's apparently have
1257a nice malloc that is well-tuned for the system.)
1258
1259=back
1260
1261=head2 Vague possibilities
1262
1263=over 4
1264
aa689395 1265=item MacPerl
1266
3e3baf6d 1267Get some of the Macintosh stuff folded back into the main distribution.
aa689395 1268
1269=item gconvert replacement
1270
1271Maybe include a replacement function that doesn't lose data in rare
1272cases of coercion between string and numerical values.
1273
aa689395 1274=item Improve makedepend
1275
1276The current makedepend process is clunky and annoyingly slow, but it
1277works for most folks. Alas, it assumes that there is a filename
1278$firstmakefile that the B<make> command will try to use before it uses
1279F<Makefile>. Such may not be the case for all B<make> commands,
1280particularly those on non-Unix systems.
1281
1282Probably some variant of the BSD F<.depend> file will be useful.
1283We ought to check how other packages do this, if they do it at all.
1284We could probably pre-generate the dependencies (with the exception of
1285malloc.o, which could probably be determined at F<Makefile.SH>
1286extraction time.
1287
1288=item GNU Makefile standard targets
1289
1290GNU software generally has standardized Makefile targets. Unless we
1291have good reason to do otherwise, I see no reason not to support them.
1292
1293=item File locking
1294
1295Somehow, straighten out, document, and implement lockf(), flock(),
1296and/or fcntl() file locking. It's a mess.
1297
1298=back
1299
fb73857a 1300=head1 AUTHORS
aa689395 1301
fb73857a 1302Original author: Andy Dougherty doughera@lafcol.lafayette.edu .
1303Additions by Chip Salzenberg chip@perl.com and
1304Tim Bunce Tim.Bunce@ig.co.uk .
aa689395 1305
1306All opinions expressed herein are those of the authorZ<>(s).
1307
1308=head1 LAST MODIFIED
1309
93341792 1310$Id: pumpkin.pod,v 1.21 1998/07/14 17:54:56 doughera Released $