This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Add "safely supporting POSIX SA_SIGINFO" to perltodo, as described in
[perl5.git] / pod / perltodo.pod
CommitLineData
7711098a
GS
1=head1 NAME
2
3perltodo - Perl TO-DO List
4
5=head1 DESCRIPTION
e50bb9a1 6
52960e22
JC
7This is a list of wishes for Perl. The tasks we think are smaller or
8easier are listed first. Anyone is welcome to work on any of these,
9but it's a good idea to first contact I<perl5-porters@perl.org> to
10avoid duplication of effort, and to learn from any previous attempts.
11By all means contact a pumpking privately first if you prefer.
e50bb9a1 12
0bdfc961
NC
13Whilst patches to make the list shorter are most welcome, ideas to add to
14the list are also encouraged. Check the perl5-porters archives for past
15ideas, and any discussion about them. One set of archives may be found at:
e50bb9a1 16
0bdfc961 17 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/
938c8732 18
617eabfa
NC
19What can we offer you in return? Fame, fortune, and everlasting glory? Maybe
20not, but if your patch is incorporated, then we'll add your name to the
21F<AUTHORS> file, which ships in the official distribution. How many other
22programming languages offer you 1 line of immortality?
938c8732 23
0bdfc961 24=head1 Tasks that only need Perl knowledge
e50bb9a1 25
5a176cbc
NC
26=head2 Remove duplication of test setup.
27
28Schwern notes, that there's duplication of code - lots and lots of tests have
29some variation on the big block of C<$Is_Foo> checks. We can safely put this
30into a file, change it to build an C<%Is> hash and require it. Maybe just put
31it into F<test.pl>. Throw in the handy tainting subroutines.
32
412f19a0
NC
33=head2 merge common code in installperl and installman
34
35There are some common subroutines and a common C<BEGIN> block in F<installperl>
36and F<installman>. These should probably be merged. It would also be good to
37check for duplication in all the utility scripts supplied in the source
38tarball. It might be good to move them all to a subdirectory, but this would
39require careful checking to find all places that call them, and change those
40correctly.
41
0bdfc961 42=head2 common test code for timed bail out
e50bb9a1 43
0bdfc961
NC
44Write portable self destruct code for tests to stop them burning CPU in
45infinite loops. This needs to avoid using alarm, as some of the tests are
46testing alarm/sleep or timers.
e50bb9a1 47
87a942b1 48=head2 POD -E<gt> HTML conversion in the core still sucks
e50bb9a1 49
938c8732 50Which is crazy given just how simple POD purports to be, and how simple HTML
adebf063
NC
51can be. It's not actually I<as> simple as it sounds, particularly with the
52flexibility POD allows for C<=item>, but it would be good to improve the
53visual appeal of the HTML generated, and to avoid it having any validation
54errors. See also L</make HTML install work>, as the layout of installation tree
55is needed to improve the cross-linking.
938c8732 56
dc0fb092
SP
57The addition of C<Pod::Simple> and its related modules may make this task
58easier to complete.
59
8537f021
RGS
60=head2 merge checkpods and podchecker
61
62F<pod/checkpods.PL> (and C<make check> in the F<pod/> subdirectory)
63implements a very basic check for pod files, but the errors it discovers
64aren't found by podchecker. Add this check to podchecker, get rid of
65checkpods and have C<make check> use podchecker.
66
aa237293
NC
67=head2 Parallel testing
68
b2e2905c 69(This probably impacts much more than the core: also the Test::Harness
02f21748
RGS
70and TAP::* modules on CPAN.)
71
aa237293
NC
72The core regression test suite is getting ever more comprehensive, which has
73the side effect that it takes longer to run. This isn't so good. Investigate
74whether it would be feasible to give the harness script the B<option> of
75running sets of tests in parallel. This would be useful for tests in
76F<t/op/*.t> and F<t/uni/*.t> and maybe some sets of tests in F<lib/>.
77
78Questions to answer
79
80=over 4
81
82=item 1
83
84How does screen layout work when you're running more than one test?
85
86=item 2
87
88How does the caller of test specify how many tests to run in parallel?
89
90=item 3
91
92How do setup/teardown tests identify themselves?
93
94=back
95
96Pugs already does parallel testing - can their approach be re-used?
97
0bdfc961 98=head2 Make Schwern poorer
e50bb9a1 99
613bd4f7 100We should have tests for everything. When all the core's modules are tested,
0bdfc961
NC
101Schwern has promised to donate to $500 to TPF. We may need volunteers to
102hold him upside down and shake vigorously in order to actually extract the
103cash.
3958b146 104
0bdfc961 105=head2 Improve the coverage of the core tests
e50bb9a1 106
02f21748
RGS
107Use Devel::Cover to ascertain the core modules's test coverage, then add
108tests that are currently missing.
30222c0f 109
0bdfc961 110=head2 test B
e50bb9a1 111
0bdfc961 112A full test suite for the B module would be nice.
e50bb9a1 113
636e63cb
NC
114=head2 Deparse inlined constants
115
116Code such as this
117
118 use constant PI => 4;
119 warn PI
120
121will currently deparse as
122
123 use constant ('PI', 4);
124 warn 4;
125
126because the tokenizer inlines the value of the constant subroutine C<PI>.
127This allows various compile time optimisations, such as constant folding
128and dead code elimination. Where these haven't happened (such as the example
129above) it ought be possible to make B::Deparse work out the name of the
130original constant, because just enough information survives in the symbol
131table to do this. Specifically, the same scalar is used for the constant in
132the optree as is used for the constant subroutine, so by iterating over all
133symbol tables and generating a mapping of SV address to constant name, it
134would be possible to provide B::Deparse with this functionality.
135
0bdfc961 136=head2 A decent benchmark
e50bb9a1 137
617eabfa 138C<perlbench> seems impervious to any recent changes made to the perl core. It
0bdfc961
NC
139would be useful to have a reasonable general benchmarking suite that roughly
140represented what current perl programs do, and measurably reported whether
141tweaks to the core improve, degrade or don't really affect performance, to
142guide people attempting to optimise the guts of perl. Gisle would welcome
143new tests for perlbench.
6168cf99 144
0bdfc961 145=head2 fix tainting bugs
6168cf99 146
0bdfc961
NC
147Fix the bugs revealed by running the test suite with the C<-t> switch (via
148C<make test.taintwarn>).
e50bb9a1 149
0bdfc961 150=head2 Dual life everything
e50bb9a1 151
0bdfc961
NC
152As part of the "dists" plan, anything that doesn't belong in the smallest perl
153distribution needs to be dual lifed. Anything else can be too. Figure out what
154changes would be needed to package that module and its tests up for CPAN, and
155do so. Test it with older perl releases, and fix the problems you find.
e50bb9a1 156
a393eb28
RGS
157To make a minimal perl distribution, it's useful to look at
158F<t/lib/commonsense.t>.
159
0bdfc961 160=head2 Improving C<threads::shared>
722d2a37 161
0bdfc961
NC
162Investigate whether C<threads::shared> could share aggregates properly with
163only Perl level changes to shared.pm
722d2a37 164
0bdfc961 165=head2 POSIX memory footprint
e50bb9a1 166
0bdfc961
NC
167Ilya observed that use POSIX; eats memory like there's no tomorrow, and at
168various times worked to cut it down. There is probably still fat to cut out -
169for example POSIX passes Exporter some very memory hungry data structures.
e50bb9a1 170
eed36644
NC
171=head2 embed.pl/makedef.pl
172
173There is a script F<embed.pl> that generates several header files to prefix
174all of Perl's symbols in a consistent way, to provide some semblance of
175namespace support in C<C>. Functions are declared in F<embed.fnc>, variables
907b3e23 176in F<interpvar.h>. Quite a few of the functions and variables
eed36644
NC
177are conditionally declared there, using C<#ifdef>. However, F<embed.pl>
178doesn't understand the C macros, so the rules about which symbols are present
179when is duplicated in F<makedef.pl>. Writing things twice is bad, m'kay.
180It would be good to teach C<embed.pl> to understand the conditional
181compilation, and hence remove the duplication, and the mistakes it has caused.
e50bb9a1 182
801de10e
NC
183=head2 use strict; and AutoLoad
184
185Currently if you write
186
187 package Whack;
188 use AutoLoader 'AUTOLOAD';
189 use strict;
190 1;
191 __END__
192 sub bloop {
193 print join (' ', No, strict, here), "!\n";
194 }
195
196then C<use strict;> isn't in force within the autoloaded subroutines. It would
197be more consistent (and less surprising) to arrange for all lexical pragmas
198in force at the __END__ block to be in force within each autoloaded subroutine.
199
773b3597
RGS
200There's a similar problem with SelfLoader.
201
0bdfc961 202=head1 Tasks that need a little sysadmin-type knowledge
e50bb9a1 203
0bdfc961
NC
204Or if you prefer, tasks that you would learn from, and broaden your skills
205base...
e50bb9a1 206
cd793d32 207=head2 make HTML install work
e50bb9a1 208
adebf063
NC
209There is an C<installhtml> target in the Makefile. It's marked as
210"experimental". It would be good to get this tested, make it work reliably, and
211remove the "experimental" tag. This would include
212
213=over 4
214
215=item 1
216
217Checking that cross linking between various parts of the documentation works.
218In particular that links work between the modules (files with POD in F<lib/>)
219and the core documentation (files in F<pod/>)
220
221=item 2
222
617eabfa
NC
223Work out how to split C<perlfunc> into chunks, preferably one per function
224group, preferably with general case code that could be used elsewhere.
225Challenges here are correctly identifying the groups of functions that go
226together, and making the right named external cross-links point to the right
227page. Things to be aware of are C<-X>, groups such as C<getpwnam> to
228C<endservent>, two or more C<=items> giving the different parameter lists, such
229as
adebf063
NC
230
231 =item substr EXPR,OFFSET,LENGTH,REPLACEMENT
adebf063 232 =item substr EXPR,OFFSET,LENGTH
adebf063
NC
233 =item substr EXPR,OFFSET
234
235and different parameter lists having different meanings. (eg C<select>)
236
237=back
3a89a73c 238
0bdfc961
NC
239=head2 compressed man pages
240
241Be able to install them. This would probably need a configure test to see how
242the system does compressed man pages (same directory/different directory?
243same filename/different filename), as well as tweaking the F<installman> script
244to compress as necessary.
245
30222c0f
NC
246=head2 Add a code coverage target to the Makefile
247
248Make it easy for anyone to run Devel::Cover on the core's tests. The steps
249to do this manually are roughly
250
251=over 4
252
253=item *
254
255do a normal C<Configure>, but include Devel::Cover as a module to install
256(see F<INSTALL> for how to do this)
257
258=item *
259
260 make perl
261
262=item *
263
264 cd t; HARNESS_PERL_SWITCHES=-MDevel::Cover ./perl -I../lib harness
265
266=item *
267
268Process the resulting Devel::Cover database
269
270=back
271
272This just give you the coverage of the F<.pm>s. To also get the C level
273coverage you need to
274
275=over 4
276
277=item *
278
279Additionally tell C<Configure> to use the appropriate C compiler flags for
280C<gcov>
281
282=item *
283
284 make perl.gcov
285
286(instead of C<make perl>)
287
288=item *
289
290After running the tests run C<gcov> to generate all the F<.gcov> files.
291(Including down in the subdirectories of F<ext/>
292
293=item *
294
295(From the top level perl directory) run C<gcov2perl> on all the C<.gcov> files
296to get their stats into the cover_db directory.
297
298=item *
299
300Then process the Devel::Cover database
301
302=back
303
304It would be good to add a single switch to C<Configure> to specify that you
305wanted to perform perl level coverage, and another to specify C level
306coverage, and have C<Configure> and the F<Makefile> do all the right things
307automatically.
308
02f21748 309=head2 Make Config.pm cope with differences between built and installed perl
0bdfc961
NC
310
311Quite often vendors ship a perl binary compiled with their (pay-for)
312compilers. People install a free compiler, such as gcc. To work out how to
313build extensions, Perl interrogates C<%Config>, so in this situation
314C<%Config> describes compilers that aren't there, and extension building
315fails. This forces people into choosing between re-compiling perl themselves
316using the compiler they have, or only using modules that the vendor ships.
317
318It would be good to find a way teach C<Config.pm> about the installation setup,
319possibly involving probing at install time or later, so that the C<%Config> in
320a binary distribution better describes the installed machine, when the
321installed machine differs from the build machine in some significant way.
322
728f4ecd
NC
323=head2 linker specification files
324
325Some platforms mandate that you provide a list of a shared library's external
326symbols to the linker, so the core already has the infrastructure in place to
327do this for generating shared perl libraries. My understanding is that the
328GNU toolchain can accept an optional linker specification file, and restrict
329visibility just to symbols declared in that file. It would be good to extend
330F<makedef.pl> to support this format, and to provide a means within
331C<Configure> to enable it. This would allow Unix users to test that the
332export list is correct, and to build a perl that does not pollute the global
333namespace with private symbols.
334
a229ae3b
RGS
335=head2 Cross-compile support
336
337Currently C<Configure> understands C<-Dusecrosscompile> option. This option
338arranges for building C<miniperl> for TARGET machine, so this C<miniperl> is
339assumed then to be copied to TARGET machine and used as a replacement of full
340C<perl> executable.
341
d1307786 342This could be done little differently. Namely C<miniperl> should be built for
a229ae3b 343HOST and then full C<perl> with extensions should be compiled for TARGET.
d1307786 344This, however, might require extra trickery for %Config: we have one config
87a942b1
JH
345first for HOST and then another for TARGET. Tools like MakeMaker will be
346mightily confused. Having around two different types of executables and
347libraries (HOST and TARGET) makes life interesting for Makefiles and
348shell (and Perl) scripts. There is $Config{run}, normally empty, which
349can be used as an execution wrapper. Also note that in some
350cross-compilation/execution environments the HOST and the TARGET do
351not see the same filesystem(s), the $Config{run} may need to do some
352file/directory copying back and forth.
0bdfc961 353
8537f021
RGS
354=head2 roffitall
355
356Make F<pod/roffitall> be updated by F<pod/buildtoc>.
357
0bdfc961
NC
358=head1 Tasks that need a little C knowledge
359
360These tasks would need a little C knowledge, but don't need any specific
361background or experience with XS, or how the Perl interpreter works
362
3d826b29
NC
363=head2 Weed out needless PERL_UNUSED_ARG
364
365The C code uses the macro C<PERL_UNUSED_ARG> to stop compilers warning about
366unused arguments. Often the arguments can't be removed, as there is an
367external constraint that determines the prototype of the function, so this
368approach is valid. However, there are some cases where C<PERL_UNUSED_ARG>
369could be removed. Specifically
370
371=over 4
372
373=item *
374
375The prototypes of (nearly all) static functions can be changed
376
377=item *
378
379Unused arguments generated by short cut macros are wasteful - the short cut
380macro used can be changed.
381
382=back
383
fbf638cb
RGS
384=head2 Modernize the order of directories in @INC
385
386The way @INC is laid out by default, one cannot upgrade core (dual-life)
387modules without overwriting files. This causes problems for binary
3d14fd97
AD
388package builders. One possible proposal is laid out in this
389message:
390L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2002-04/msg02380.html>.
fbf638cb 391
bcbaa2d5
RGS
392=head2 -Duse32bit*
393
394Natively 64-bit systems need neither -Duse64bitint nor -Duse64bitall.
395On these systems, it might be the default compilation mode, and there
396is currently no guarantee that passing no use64bitall option to the
397Configure process will build a 32bit perl. Implementing -Duse32bit*
398options would be nice for perl 5.12.
399
0bdfc961 400=head2 Make it clear from -v if this is the exact official release
89007cb3 401
617eabfa
NC
402Currently perl from C<p4>/C<rsync> ships with a F<patchlevel.h> file that
403usually defines one local patch, of the form "MAINT12345" or "RC1". The output
404of perl -v doesn't report that a perl isn't an official release, and this
89007cb3 405information can get lost in bugs reports. Because of this, the minor version
fa11829f 406isn't bumped up until RC time, to minimise the possibility of versions of perl
89007cb3
NC
407escaping that believe themselves to be newer than they actually are.
408
409It would be useful to find an elegant way to have the "this is an interim
410maintenance release" or "this is a release candidate" in the terse -v output,
411and have it so that it's easy for the pumpking to remove this just as the
412release tarball is rolled up. This way the version pulled out of rsync would
413always say "I'm a development release" and it would be safe to bump the
414reported minor version as soon as a release ships, which would aid perl
415developers.
416
0bdfc961
NC
417This task is really about thinking of an elegant way to arrange the C source
418such that it's trivial for the Pumpking to flag "this is an official release"
419when making a tarball, yet leave the default source saying "I'm not the
420official release".
421
fee0a0f7 422=head2 Profile Perl - am I hot or not?
62403a3c 423
fee0a0f7
NC
424The Perl source code is stable enough that it makes sense to profile it,
425identify and optimise the hotspots. It would be good to measure the
426performance of the Perl interpreter using free tools such as cachegrind,
427gprof, and dtrace, and work to reduce the bottlenecks they reveal.
428
429As part of this, the idea of F<pp_hot.c> is that it contains the I<hot> ops,
430the ops that are most commonly used. The idea is that by grouping them, their
431object code will be adjacent in the executable, so they have a greater chance
432of already being in the CPU cache (or swapped in) due to being near another op
433already in use.
62403a3c
NC
434
435Except that it's not clear if these really are the most commonly used ops. So
fee0a0f7
NC
436as part of exercising your skills with coverage and profiling tools you might
437want to determine what ops I<really> are the most commonly used. And in turn
438suggest evictions and promotions to achieve a better F<pp_hot.c>.
62403a3c 439
98fed0ad
NC
440=head2 Allocate OPs from arenas
441
442Currently all new OP structures are individually malloc()ed and free()d.
443All C<malloc> implementations have space overheads, and are now as fast as
444custom allocates so it would both use less memory and less CPU to allocate
445the various OP structures from arenas. The SV arena code can probably be
446re-used for this.
447
539f2c54
JC
448Note that Configuring perl with C<-Accflags=-DPL_OP_SLAB_ALLOC> will use
449Perl_Slab_alloc() to pack optrees into a contiguous block, which is
450probably superior to the use of OP arenas, esp. from a cache locality
451standpoint. See L<Profile Perl - am I hot or not?>.
452
a229ae3b 453=head2 Improve win32/wince.c
0bdfc961 454
a229ae3b 455Currently, numerous functions look virtually, if not completely,
02f21748 456identical in both C<win32/wince.c> and C<win32/win32.c> files, which can't
6d71adcd
NC
457be good.
458
c5b31784
SH
459=head2 Use secure CRT functions when building with VC8 on Win32
460
461Visual C++ 2005 (VC++ 8.x) deprecated a number of CRT functions on the basis
462that they were "unsafe" and introduced differently named secure versions of
463them as replacements, e.g. instead of writing
464
465 FILE* f = fopen(__FILE__, "r");
466
467one should now write
468
469 FILE* f;
470 errno_t err = fopen_s(&f, __FILE__, "r");
471
472Currently, the warnings about these deprecations have been disabled by adding
473-D_CRT_SECURE_NO_DEPRECATE to the CFLAGS. It would be nice to remove that
474warning suppressant and actually make use of the new secure CRT functions.
475
476There is also a similar issue with POSIX CRT function names like fileno having
477been deprecated in favour of ISO C++ conformant names like _fileno. These
26a6faa8 478warnings are also currently suppressed by adding -D_CRT_NONSTDC_NO_DEPRECATE. It
c5b31784
SH
479might be nice to do as Microsoft suggest here too, although, unlike the secure
480functions issue, there is presumably little or no benefit in this case.
481
038ae9a4
SH
482=head2 Fix POSIX::access() and chdir() on Win32
483
484These functions currently take no account of DACLs and therefore do not behave
485correctly in situations where access is restricted by DACLs (as opposed to the
486read-only attribute).
487
488Furthermore, POSIX::access() behaves differently for directories having the
489read-only attribute set depending on what CRT library is being used. For
490example, the _access() function in the VC6 and VC7 CRTs (wrongly) claim that
491such directories are not writable, whereas in fact all directories are writable
492unless access is denied by DACLs. (In the case of directories, the read-only
493attribute actually only means that the directory cannot be deleted.) This CRT
494bug is fixed in the VC8 and VC9 CRTs (but, of course, the directory may still
495not actually be writable if access is indeed denied by DACLs).
496
497For the chdir() issue, see ActiveState bug #74552:
498http://bugs.activestate.com/show_bug.cgi?id=74552
499
500Therefore, DACLs should be checked both for consistency across CRTs and for
501the correct answer.
502
503(Note that perl's -w operator should not be modified to check DACLs. It has
504been written so that it reflects the state of the read-only attribute, even
505for directories (whatever CRT is being used), for symmetry with chmod().)
506
16815324
NC
507=head2 strcat(), strcpy(), strncat(), strncpy(), sprintf(), vsprintf()
508
509Maybe create a utility that checks after each libperl.a creation that
510none of the above (nor sprintf(), vsprintf(), or *SHUDDER* gets())
511ever creep back to libperl.a.
512
513 nm libperl.a | ./miniperl -alne '$o = $F[0] if /:$/; print "$o $F[1]" if $F[0] eq "U" && $F[1] =~ /^(?:strn?c(?:at|py)|v?sprintf|gets)$/'
514
515Note, of course, that this will only tell whether B<your> platform
516is using those naughty interfaces.
517
de96509d
JH
518=head2 -D_FORTIFY_SOURCE=2, -fstack-protector
519
520Recent glibcs support C<-D_FORTIFY_SOURCE=2> and recent gcc
521(4.1 onwards?) supports C<-fstack-protector>, both of which give
522protection against various kinds of buffer overflow problems.
523These should probably be used for compiling Perl whenever available,
524Configure and/or hints files should be adjusted to probe for the
525availability of these features and enable them as appropriate.
16815324 526
8964cfe0
NC
527=head2 Arenas for GPs? For MAGIC?
528
529C<struct gp> and C<struct magic> are both currently allocated by C<malloc>.
530It might be a speed or memory saving to change to using arenas. Or it might
531not. It would need some suitable benchmarking first. In particular, C<GP>s
532can probably be changed with minimal compatibility impact (probably nothing
533outside of the core, or even outside of F<gv.c> allocates them), but they
534probably aren't allocated/deallocated often enough for a speed saving. Whereas
535C<MAGIC> is allocated/deallocated more often, but in turn, is also something
536more externally visible, so changing the rules here may bite external code.
537
538
6d71adcd
NC
539=head1 Tasks that need a knowledge of XS
540
541These tasks would need C knowledge, and roughly the level of knowledge of
542the perl API that comes from writing modules that use XS to interface to
543C.
544
5d96f598
NC
545=head2 safely supporting POSIX SA_SIGINFO
546
547Some years ago Jarkko supplied patches to provide support for the POSIX
548SA_SIGINFO feature in Perl, passing the extra data to the Perl signal handler.
549
550Unfortunately, it only works with "unsafe" signals, because under safe
551signals, by the time Perl gets to run the signal handler, the extra
552information has been lost. Moreover, it's not easy to store it somewhere,
553as you can't call mutexs, or do anything else fancy, from inside a signal
554handler.
555
556So it strikes me that we could provide safe SA_SIGINFO support
557
558=over 4
559
560=item 1
561
562Provide global variables for two file descriptors
563
564=item 2
565
566When the first request is made via C<sigaction> for C<SA_SIGINFO>, create a
567pipe, store the reader in one, the writer in the other
568
569=item 3
570
571In the "safe" signal handler (C<Perl_csighandler()>/C<S_raise_signal()>), if
572the C<siginfo_t> pointer non-C<NULL>, and the writer file handle is open,
573
574=over 8
575
576=item 1
577
578serialise signal number, C<struct siginfo_t> (or at least the parts we care
579about) into a small auto char buff
580
581=item 2
582
583C<write()> that (non-blocking) to the writer fd
584
585=over 12
586
587=item 1
588
589if it writes 100%, flag the signal in a counter of "signals on the pipe" akin
590to the current per-signal-number counts
591
592=item 2
593
594if it writes 0%, assume the pipe is full. Flag the data as lost?
595
596=item 3
597
598if it writes partially, croak a panic, as your OS is broken.
599
600=back
601
602=back
603
604=item 4
605
606in the regular C<PERL_ASYNC_CHECK()> processing, if there are "signals on
607the pipe", read the data out, deserialise, build the Perl structures on
608the stack (code in C<Perl_sighandler()>, the "unsafe" handler), and call as
609usual.
610
611=back
612
613I think that this gets us decent C<SA_SIGINFO> support, without the current risk
614of running Perl code inside the signal handler context. (With all the dangers
615of things like C<malloc> corruption that that currently offers us)
616
617For more information see the thread starting with this message:
618http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-03/msg00305.html
619
6d71adcd
NC
620=head2 autovivification
621
622Make all autovivification consistent w.r.t LVALUE/RVALUE and strict/no strict;
623
624This task is incremental - even a little bit of work on it will help.
625
626=head2 Unicode in Filenames
627
628chdir, chmod, chown, chroot, exec, glob, link, lstat, mkdir, open,
629opendir, qx, readdir, readlink, rename, rmdir, stat, symlink, sysopen,
630system, truncate, unlink, utime, -X. All these could potentially accept
631Unicode filenames either as input or output (and in the case of system
632and qx Unicode in general, as input or output to/from the shell).
633Whether a filesystem - an operating system pair understands Unicode in
634filenames varies.
635
636Known combinations that have some level of understanding include
637Microsoft NTFS, Apple HFS+ (In Mac OS 9 and X) and Apple UFS (in Mac
638OS X), NFS v4 is rumored to be Unicode, and of course Plan 9. How to
639create Unicode filenames, what forms of Unicode are accepted and used
640(UCS-2, UTF-16, UTF-8), what (if any) is the normalization form used,
641and so on, varies. Finding the right level of interfacing to Perl
642requires some thought. Remember that an OS does not implicate a
643filesystem.
644
645(The Windows -C command flag "wide API support" has been at least
646temporarily retired in 5.8.1, and the -C has been repurposed, see
647L<perlrun>.)
648
87a942b1
JH
649Most probably the right way to do this would be this:
650L</"Virtualize operating system access">.
651
6d71adcd
NC
652=head2 Unicode in %ENV
653
654Currently the %ENV entries are always byte strings.
87a942b1 655See L</"Virtualize operating system access">.
6d71adcd 656
1f2e7916
JD
657=head2 Unicode and glob()
658
659Currently glob patterns and filenames returned from File::Glob::glob()
87a942b1 660are always byte strings. See L</"Virtualize operating system access">.
1f2e7916 661
dbb0c492
RGS
662=head2 Unicode and lc/uc operators
663
664Some built-in operators (C<lc>, C<uc>, etc.) behave differently, based on
665what the internal encoding of their argument is. That should not be the
666case. Maybe add a pragma to switch behaviour.
667
6d71adcd
NC
668=head2 use less 'memory'
669
670Investigate trade offs to switch out perl's choices on memory usage.
671Particularly perl should be able to give memory back.
672
673This task is incremental - even a little bit of work on it will help.
674
675=head2 Re-implement C<:unique> in a way that is actually thread-safe
676
677The old implementation made bad assumptions on several levels. A good 90%
678solution might be just to make C<:unique> work to share the string buffer
679of SvPVs. That way large constant strings can be shared between ithreads,
680such as the configuration information in F<Config>.
681
682=head2 Make tainting consistent
683
684Tainting would be easier to use if it didn't take documented shortcuts and
685allow taint to "leak" everywhere within an expression.
686
687=head2 readpipe(LIST)
688
689system() accepts a LIST syntax (and a PROGRAM LIST syntax) to avoid
690running a shell. readpipe() (the function behind qx//) could be similarly
691extended.
692
6d71adcd
NC
693=head2 Audit the code for destruction ordering assumptions
694
695Change 25773 notes
696
697 /* Need to check SvMAGICAL, as during global destruction it may be that
698 AvARYLEN(av) has been freed before av, and hence the SvANY() pointer
699 is now part of the linked list of SV heads, rather than pointing to
700 the original body. */
701 /* FIXME - audit the code for other bugs like this one. */
702
703adding the C<SvMAGICAL> check to
704
705 if (AvARYLEN(av) && SvMAGICAL(AvARYLEN(av))) {
706 MAGIC *mg = mg_find (AvARYLEN(av), PERL_MAGIC_arylen);
707
708Go through the core and look for similar assumptions that SVs have particular
709types, as all bets are off during global destruction.
710
749904bf
JH
711=head2 Extend PerlIO and PerlIO::Scalar
712
713PerlIO::Scalar doesn't know how to truncate(). Implementing this
714would require extending the PerlIO vtable.
715
716Similarly the PerlIO vtable doesn't know about formats (write()), or
717about stat(), or chmod()/chown(), utime(), or flock().
718
719(For PerlIO::Scalar it's hard to see what e.g. mode bits or ownership
720would mean.)
721
722PerlIO doesn't do directories or symlinks, either: mkdir(), rmdir(),
723opendir(), closedir(), seekdir(), rewinddir(), glob(); symlink(),
724readlink().
725
94da6c29
JH
726See also L</"Virtualize operating system access">.
727
3236f110
NC
728=head2 -C on the #! line
729
730It should be possible to make -C work correctly if found on the #! line,
731given that all perl command line options are strict ASCII, and -C changes
732only the interpretation of non-ASCII characters, and not for the script file
733handle. To make it work needs some investigation of the ordering of function
734calls during startup, and (by implication) a bit of tweaking of that order.
735
d6c1e11f
JH
736=head2 Organize error messages
737
738Perl's diagnostics (error messages, see L<perldiag>) could use
a8d0aeb9 739reorganizing and formalizing so that each error message has its
d6c1e11f
JH
740stable-for-all-eternity unique id, categorized by severity, type, and
741subsystem. (The error messages would be listed in a datafile outside
c4bd451b
CB
742of the Perl source code, and the source code would only refer to the
743messages by the id.) This clean-up and regularizing should apply
d6c1e11f
JH
744for all croak() messages.
745
746This would enable all sorts of things: easier translation/localization
747of the messages (though please do keep in mind the caveats of
748L<Locale::Maketext> about too straightforward approaches to
749translation), filtering by severity, and instead of grepping for a
750particular error message one could look for a stable error id. (Of
751course, changing the error messages by default would break all the
752existing software depending on some particular error message...)
753
754This kind of functionality is known as I<message catalogs>. Look for
755inspiration for example in the catgets() system, possibly even use it
756if available-- but B<only> if available, all platforms will B<not>
de96509d 757have catgets().
d6c1e11f
JH
758
759For the really pure at heart, consider extending this item to cover
760also the warning messages (see L<perllexwarn>, C<warnings.pl>).
3236f110 761
0bdfc961 762=head1 Tasks that need a knowledge of the interpreter
3298bd4d 763
0bdfc961
NC
764These tasks would need C knowledge, and knowledge of how the interpreter works,
765or a willingness to learn.
3298bd4d 766
718140ec
NC
767=head2 lexicals used only once
768
769This warns:
770
771 $ perl -we '$pie = 42'
772 Name "main::pie" used only once: possible typo at -e line 1.
773
774This does not:
775
776 $ perl -we 'my $pie = 42'
777
778Logically all lexicals used only once should warn, if the user asks for
d6f4ea2e
SP
779warnings. An unworked RT ticket (#5087) has been open for almost seven
780years for this discrepancy.
718140ec 781
a3d15f9a
RGS
782=head2 UTF-8 revamp
783
784The handling of Unicode is unclean in many places. For example, the regexp
785engine matches in Unicode semantics whenever the string or the pattern is
786flagged as UTF-8, but that should not be dependent on an internal storage
787detail of the string. Likewise, case folding behaviour is dependent on the
788UTF8 internal flag being on or off.
789
790=head2 Properly Unicode safe tokeniser and pads.
791
792The tokeniser isn't actually very UTF-8 clean. C<use utf8;> is a hack -
793variable names are stored in stashes as raw bytes, without the utf-8 flag
794set. The pad API only takes a C<char *> pointer, so that's all bytes too. The
795tokeniser ignores the UTF-8-ness of C<PL_rsfp>, or any SVs returned from
796source filters. All this could be fixed.
797
636e63cb
NC
798=head2 state variable initialization in list context
799
800Currently this is illegal:
801
802 state ($a, $b) = foo();
803
a2874905 804In Perl 6, C<state ($a) = foo();> and C<(state $a) = foo();> have different
a8d0aeb9 805semantics, which is tricky to implement in Perl 5 as currently they produce
a2874905 806the same opcode trees. The Perl 6 design is firm, so it would be good to
a8d0aeb9 807implement the necessary code in Perl 5. There are comments in
a2874905
NC
808C<Perl_newASSIGNOP()> that show the code paths taken by various assignment
809constructions involving state variables.
636e63cb 810
4fedb12c
RGS
811=head2 Implement $value ~~ 0 .. $range
812
813It would be nice to extend the syntax of the C<~~> operator to also
814understand numeric (and maybe alphanumeric) ranges.
a393eb28
RGS
815
816=head2 A does() built-in
817
818Like ref(), only useful. It would call the C<DOES> method on objects; it
819would also tell whether something can be dereferenced as an
820array/hash/etc., or used as a regexp, etc.
821L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-03/msg00481.html>
822
823=head2 Tied filehandles and write() don't mix
824
825There is no method on tied filehandles to allow them to be called back by
826formats.
4fedb12c 827
d10fc472 828=head2 Attach/detach debugger from running program
1626a787 829
cd793d32
NC
830The old perltodo notes "With C<gdb>, you can attach the debugger to a running
831program if you pass the process ID. It would be good to do this with the Perl
0bdfc961
NC
832debugger on a running Perl program, although I'm not sure how it would be
833done." ssh and screen do this with named pipes in /tmp. Maybe we can too.
1626a787 834
a8cb5b9e
RGS
835=head2 Optimize away empty destructors
836
837Defining an empty DESTROY method might be useful (notably in
838AUTOLOAD-enabled classes), but it's still a bit expensive to call. That
839could probably be optimized.
840
0bdfc961
NC
841=head2 LVALUE functions for lists
842
843The old perltodo notes that lvalue functions don't work for list or hash
844slices. This would be good to fix.
845
846=head2 LVALUE functions in the debugger
847
848The old perltodo notes that lvalue functions don't work in the debugger. This
849would be good to fix.
850
0bdfc961
NC
851=head2 regexp optimiser optional
852
853The regexp optimiser is not optional. It should configurable to be, to allow
854its performance to be measured, and its bugs to be easily demonstrated.
855
02f21748
RGS
856=head2 delete &function
857
858Allow to delete functions. One can already undef them, but they're still
859in the stash.
860
ef36c6a7
RGS
861=head2 C</w> regex modifier
862
863That flag would enable to match whole words, and also to interpolate
864arrays as alternations. With it, C</P/w> would be roughly equivalent to:
865
866 do { local $"='|'; /\b(?:P)\b/ }
867
868See L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-01/msg00400.html>
869for the discussion.
870
0bdfc961
NC
871=head2 optional optimizer
872
873Make the peephole optimizer optional. Currently it performs two tasks as
874it walks the optree - genuine peephole optimisations, and necessary fixups of
875ops. It would be good to find an efficient way to switch out the
876optimisations whilst keeping the fixups.
877
878=head2 You WANT *how* many
879
880Currently contexts are void, scalar and list. split has a special mechanism in
881place to pass in the number of return values wanted. It would be useful to
882have a general mechanism for this, backwards compatible and little speed hit.
883This would allow proposals such as short circuiting sort to be implemented
884as a module on CPAN.
885
886=head2 lexical aliases
887
888Allow lexical aliases (maybe via the syntax C<my \$alias = \$foo>.
889
890=head2 entersub XS vs Perl
891
892At the moment pp_entersub is huge, and has code to deal with entering both
893perl and XS subroutines. Subroutine implementations rarely change between
894perl and XS at run time, so investigate using 2 ops to enter subs (one for
895XS, one for perl) and swap between if a sub is redefined.
2810d901 896
de535794 897=head2 Self-ties
2810d901 898
de535794 899Self-ties are currently illegal because they caused too many segfaults. Maybe
a8d0aeb9 900the causes of these could be tracked down and self-ties on all types
de535794 901reinstated.
0bdfc961
NC
902
903=head2 Optimize away @_
904
905The old perltodo notes "Look at the "reification" code in C<av.c>".
906
f092b1f4
RGS
907=head2 The yada yada yada operators
908
909Perl 6's Synopsis 3 says:
910
911I<The ... operator is the "yada, yada, yada" list operator, which is used as
912the body in function prototypes. It complains bitterly (by calling fail)
913if it is ever executed. Variant ??? calls warn, and !!! calls die.>
914
915Those would be nice to add to Perl 5. That could be done without new ops.
916
87a942b1
JH
917=head2 Virtualize operating system access
918
919Implement a set of "vtables" that virtualizes operating system access
920(open(), mkdir(), unlink(), readdir(), getenv(), etc.) At the very
921least these interfaces should take SVs as "name" arguments instead of
922bare char pointers; probably the most flexible and extensible way
e1a3d5d1
JH
923would be for the Perl-facing interfaces to accept HVs. The system
924needs to be per-operating-system and per-file-system
925hookable/filterable, preferably both from XS and Perl level
87a942b1
JH
926(L<perlport/"Files and Filesystems"> is good reading at this point,
927in fact, all of L<perlport> is.)
928
e1a3d5d1
JH
929This has actually already been implemented (but only for Win32),
930take a look at F<iperlsys.h> and F<win32/perlhost.h>. While all Win32
931variants go through a set of "vtables" for operating system access,
932non-Win32 systems currently go straight for the POSIX/UNIX-style
933system/library call. Similar system as for Win32 should be
934implemented for all platforms. The existing Win32 implementation
935probably does not need to survive alongside this proposed new
936implementation, the approaches could be merged.
87a942b1
JH
937
938What would this give us? One often-asked-for feature this would
94da6c29
JH
939enable is using Unicode for filenames, and other "names" like %ENV,
940usernames, hostnames, and so forth.
941(See L<perlunicode/"When Unicode Does Not Happen">.)
942
943But this kind of virtualization would also allow for things like
944virtual filesystems, virtual networks, and "sandboxes" (though as long
945as dynamic loading of random object code is allowed, not very safe
946sandboxes since external code of course know not of Perl's vtables).
947An example of a smaller "sandbox" is that this feature can be used to
948implement per-thread working directories: Win32 already does this.
949
950See also L</"Extend PerlIO and PerlIO::Scalar">.
87a942b1 951
ac6197af
NC
952=head2 Investigate PADTMP hash pessimisation
953
954The peephole optimier converts constants used for hash key lookups to shared
057163d7 955hash key scalars. Under ithreads, something is undoing this work.
ac6197af
NC
956See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-09/msg00793.html
957
057163d7
NC
958=head2 Store the current pad in the OP slab allocator
959
960=for clarification
961I hope that I got that "current pad" part correct
962
963Currently we leak ops in various cases of parse failure. I suggested that we
964could solve this by always using the op slab allocator, and walking it to
965free ops. Dave comments that as some ops are already freed during optree
966creation one would have to mark which ops are freed, and not double free them
967when walking the slab. He notes that one problem with this is that for some ops
968you have to know which pad was current at the time of allocation, which does
969change. I suggested storing a pointer to the current pad in the memory allocated
970for the slab, and swapping to a new slab each time the pad changes. Dave thinks
971that this would work.
972
52960e22
JC
973=head2 repack the optree
974
975Repacking the optree after execution order is determined could allow
057163d7
NC
976removal of NULL ops, and optimal ordering of OPs with respect to cache-line
977filling. The slab allocator could be reused for this purpose. I think that
978the best way to do this is to make it an optional step just before the
979completed optree is attached to anything else, and to use the slab allocator
980unchanged, so that freeing ops is identical whether or not this step runs.
981Note that the slab allocator allocates ops downwards in memory, so one would
982have to actually "allocate" the ops in reverse-execution order to get them
983contiguous in memory in execution order.
984
985See http://www.nntp.perl.org/group/perl.perl5.porters/2007/12/msg131975.html
986
987Note that running this copy, and then freeing all the old location ops would
988cause their slabs to be freed, which would eliminate possible memory wastage if
989the previous suggestion is implemented, and we swap slabs more frequently.
52960e22 990
12e06b6f
NC
991=head2 eliminate incorrect line numbers in warnings
992
993This code
994
995 use warnings;
996 my $undef;
997
998 if ($undef == 3) {
999 } elsif ($undef == 0) {
1000 }
1001
18a16cc5 1002used to produce this output:
12e06b6f
NC
1003
1004 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1005 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1006
18a16cc5
NC
1007where the line of the second warning was misreported - it should be line 5.
1008Rafael fixed this - the problem arose because there was no nextstate OP
1009between the execution of the C<if> and the C<elsif>, hence C<PL_curcop> still
1010reports that the currently executing line is line 4. The solution was to inject
1011a nextstate OPs for each C<elsif>, although it turned out that the nextstate
1012OP needed to be a nulled OP, rather than a live nextstate OP, else other line
1013numbers became misreported. (Jenga!)
12e06b6f
NC
1014
1015The problem is more general than C<elsif> (although the C<elsif> case is the
1016most common and the most confusing). Ideally this code
1017
1018 use warnings;
1019 my $undef;
1020
1021 my $a = $undef + 1;
1022 my $b
1023 = $undef
1024 + 1;
1025
1026would produce this output
1027
1028 Use of uninitialized value $undef in addition (+) at wrong.pl line 4.
1029 Use of uninitialized value $undef in addition (+) at wrong.pl line 7.
1030
1031(rather than lines 4 and 5), but this would seem to require every OP to carry
1032(at least) line number information.
1033
1034What might work is to have an optional line number in memory just before the
1035BASEOP structure, with a flag bit in the op to say whether it's present.
1036Initially during compile every OP would carry its line number. Then add a late
1037pass to the optimiser (potentially combined with L</repack the optree>) which
1038looks at the two ops on every edge of the graph of the execution path. If
1039the line number changes, flags the destination OP with this information.
1040Once all paths are traced, replace every op with the flag with a
1041nextstate-light op (that just updates C<PL_curcop>), which in turn then passes
1042control on to the true op. All ops would then be replaced by variants that
1043do not store the line number. (Which, logically, why it would work best in
1044conjunction with L</repack the optree>, as that is already copying/reallocating
1045all the OPs)
1046
18a16cc5
NC
1047(Although I should note that we're not certain that doing this for the general
1048case is worth it)
1049
52960e22
JC
1050=head2 optimize tail-calls
1051
1052Tail-calls present an opportunity for broadly applicable optimization;
1053anywhere that C<< return foo(...) >> is called, the outer return can
1054be replaced by a goto, and foo will return directly to the outer
1055caller, saving (conservatively) 25% of perl's call&return cost, which
1056is relatively higher than in C. The scheme language is known to do
1057this heavily. B::Concise provides good insight into where this
1058optimization is possible, ie anywhere entersub,leavesub op-sequence
1059occurs.
1060
1061 perl -MO=Concise,-exec,a,b,-main -e 'sub a{ 1 }; sub b {a()}; b(2)'
1062
1063Bottom line on this is probably a new pp_tailcall function which
1064combines the code in pp_entersub, pp_leavesub. This should probably
1065be done 1st in XS, and using B::Generate to patch the new OP into the
1066optrees.
1067
0bdfc961
NC
1068=head1 Big projects
1069
1070Tasks that will get your name mentioned in the description of the "Highlights
87a942b1 1071of 5.12"
0bdfc961
NC
1072
1073=head2 make ithreads more robust
1074
4e577f8b 1075Generally make ithreads more robust. See also L</iCOW>
0bdfc961
NC
1076
1077This task is incremental - even a little bit of work on it will help, and
1078will be greatly appreciated.
1079
6c047da7
YST
1080One bit would be to write the missing code in sv.c:Perl_dirp_dup.
1081
59c7f7d5
RGS
1082Fix Perl_sv_dup, et al so that threads can return objects.
1083
0bdfc961
NC
1084=head2 iCOW
1085
1086Sarathy and Arthur have a proposal for an improved Copy On Write which
1087specifically will be able to COW new ithreads. If this can be implemented
1088it would be a good thing.
1089
1090=head2 (?{...}) closures in regexps
1091
1092Fix (or rewrite) the implementation of the C</(?{...})/> closures.
1093
1094=head2 A re-entrant regexp engine
1095
1096This will allow the use of a regex from inside (?{ }), (??{ }) and
1097(?(?{ })|) constructs.
6bda09f9 1098
6bda09f9
YO
1099=head2 Add class set operations to regexp engine
1100
1101Apparently these are quite useful. Anyway, Jeffery Friedl wants them.
1102
1103demerphq has this on his todo list, but right at the bottom.