This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Note the task TODO of splitting the linker from the compiler.
[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<$Is_Foo> checks. We can safely put this
30into a file, change it to build an C<%Is> hash and require it. Maybe just put
31it into F<test.pl>. Throw in the handy tainting subroutines.
32
87a942b1 33=head2 POD -E<gt> HTML conversion in the core still sucks
e50bb9a1 34
938c8732 35Which is crazy given just how simple POD purports to be, and how simple HTML
adebf063
NC
36can be. It's not actually I<as> simple as it sounds, particularly with the
37flexibility POD allows for C<=item>, but it would be good to improve the
38visual appeal of the HTML generated, and to avoid it having any validation
39errors. See also L</make HTML install work>, as the layout of installation tree
40is needed to improve the cross-linking.
938c8732 41
dc0fb092
SP
42The addition of C<Pod::Simple> and its related modules may make this task
43easier to complete.
44
8537f021
RGS
45=head2 merge checkpods and podchecker
46
47F<pod/checkpods.PL> (and C<make check> in the F<pod/> subdirectory)
48implements a very basic check for pod files, but the errors it discovers
49aren't found by podchecker. Add this check to podchecker, get rid of
50checkpods and have C<make check> use podchecker.
51
aa237293
NC
52=head2 Parallel testing
53
b2e2905c 54(This probably impacts much more than the core: also the Test::Harness
02f21748
RGS
55and TAP::* modules on CPAN.)
56
c707cc00
NC
57All of the tests in F<t/> can now be run in parallel, if C<$ENV{TEST_JOBS}>
58is set. However, tests within each directory in F<ext> and F<lib> are still
59run in series, with directories run in parallel. This is an adequate
60heuristic, but it might be possible to relax it further, and get more
61throughput. Specifically, it would be good to audit all of F<lib/*.t>, and
62make them use C<File::Temp>.
aa237293 63
0bdfc961 64=head2 Make Schwern poorer
e50bb9a1 65
613bd4f7 66We should have tests for everything. When all the core's modules are tested,
0bdfc961
NC
67Schwern has promised to donate to $500 to TPF. We may need volunteers to
68hold him upside down and shake vigorously in order to actually extract the
69cash.
3958b146 70
0bdfc961 71=head2 Improve the coverage of the core tests
e50bb9a1 72
02f21748
RGS
73Use Devel::Cover to ascertain the core modules's test coverage, then add
74tests that are currently missing.
30222c0f 75
0bdfc961 76=head2 test B
e50bb9a1 77
0bdfc961 78A full test suite for the B module would be nice.
e50bb9a1 79
636e63cb
NC
80=head2 Deparse inlined constants
81
82Code such as this
83
84 use constant PI => 4;
85 warn PI
86
87will currently deparse as
88
89 use constant ('PI', 4);
90 warn 4;
91
92because the tokenizer inlines the value of the constant subroutine C<PI>.
93This allows various compile time optimisations, such as constant folding
94and dead code elimination. Where these haven't happened (such as the example
95above) it ought be possible to make B::Deparse work out the name of the
96original constant, because just enough information survives in the symbol
97table to do this. Specifically, the same scalar is used for the constant in
98the optree as is used for the constant subroutine, so by iterating over all
99symbol tables and generating a mapping of SV address to constant name, it
100would be possible to provide B::Deparse with this functionality.
101
0bdfc961 102=head2 A decent benchmark
e50bb9a1 103
617eabfa 104C<perlbench> seems impervious to any recent changes made to the perl core. It
0bdfc961
NC
105would be useful to have a reasonable general benchmarking suite that roughly
106represented what current perl programs do, and measurably reported whether
107tweaks to the core improve, degrade or don't really affect performance, to
108guide people attempting to optimise the guts of perl. Gisle would welcome
109new tests for perlbench.
6168cf99 110
0bdfc961 111=head2 fix tainting bugs
6168cf99 112
0bdfc961
NC
113Fix the bugs revealed by running the test suite with the C<-t> switch (via
114C<make test.taintwarn>).
e50bb9a1 115
0bdfc961 116=head2 Dual life everything
e50bb9a1 117
0bdfc961
NC
118As part of the "dists" plan, anything that doesn't belong in the smallest perl
119distribution needs to be dual lifed. Anything else can be too. Figure out what
120changes would be needed to package that module and its tests up for CPAN, and
121do so. Test it with older perl releases, and fix the problems you find.
e50bb9a1 122
a393eb28
RGS
123To make a minimal perl distribution, it's useful to look at
124F<t/lib/commonsense.t>.
125
c2aba5b8
RGS
126=head2 Bundle dual life modules in ext/
127
128For maintenance (and branch merging) reasons, it would be useful to move
129some architecture-independent dual-life modules from lib/ to ext/, if this
130has no negative impact on the build of perl itself.
131
0bdfc961 132=head2 POSIX memory footprint
e50bb9a1 133
0bdfc961
NC
134Ilya observed that use POSIX; eats memory like there's no tomorrow, and at
135various times worked to cut it down. There is probably still fat to cut out -
136for example POSIX passes Exporter some very memory hungry data structures.
e50bb9a1 137
eed36644
NC
138=head2 embed.pl/makedef.pl
139
140There is a script F<embed.pl> that generates several header files to prefix
141all of Perl's symbols in a consistent way, to provide some semblance of
142namespace support in C<C>. Functions are declared in F<embed.fnc>, variables
907b3e23 143in F<interpvar.h>. Quite a few of the functions and variables
eed36644
NC
144are conditionally declared there, using C<#ifdef>. However, F<embed.pl>
145doesn't understand the C macros, so the rules about which symbols are present
146when is duplicated in F<makedef.pl>. Writing things twice is bad, m'kay.
147It would be good to teach C<embed.pl> to understand the conditional
148compilation, and hence remove the duplication, and the mistakes it has caused.
e50bb9a1 149
801de10e
NC
150=head2 use strict; and AutoLoad
151
152Currently if you write
153
154 package Whack;
155 use AutoLoader 'AUTOLOAD';
156 use strict;
157 1;
158 __END__
159 sub bloop {
160 print join (' ', No, strict, here), "!\n";
161 }
162
163then C<use strict;> isn't in force within the autoloaded subroutines. It would
164be more consistent (and less surprising) to arrange for all lexical pragmas
165in force at the __END__ block to be in force within each autoloaded subroutine.
166
773b3597
RGS
167There's a similar problem with SelfLoader.
168
91d0cbf6
NC
169=head2 profile installman
170
171The F<installman> script is slow. All it is doing text processing, which we're
172told is something Perl is good at. So it would be nice to know what it is doing
173that is taking so much CPU, and where possible address it.
174
175
0bdfc961 176=head1 Tasks that need a little sysadmin-type knowledge
e50bb9a1 177
0bdfc961
NC
178Or if you prefer, tasks that you would learn from, and broaden your skills
179base...
e50bb9a1 180
cd793d32 181=head2 make HTML install work
e50bb9a1 182
adebf063
NC
183There is an C<installhtml> target in the Makefile. It's marked as
184"experimental". It would be good to get this tested, make it work reliably, and
185remove the "experimental" tag. This would include
186
187=over 4
188
189=item 1
190
191Checking that cross linking between various parts of the documentation works.
192In particular that links work between the modules (files with POD in F<lib/>)
193and the core documentation (files in F<pod/>)
194
195=item 2
196
617eabfa
NC
197Work out how to split C<perlfunc> into chunks, preferably one per function
198group, preferably with general case code that could be used elsewhere.
199Challenges here are correctly identifying the groups of functions that go
200together, and making the right named external cross-links point to the right
201page. Things to be aware of are C<-X>, groups such as C<getpwnam> to
202C<endservent>, two or more C<=items> giving the different parameter lists, such
203as
adebf063
NC
204
205 =item substr EXPR,OFFSET,LENGTH,REPLACEMENT
adebf063 206 =item substr EXPR,OFFSET,LENGTH
adebf063
NC
207 =item substr EXPR,OFFSET
208
209and different parameter lists having different meanings. (eg C<select>)
210
211=back
3a89a73c 212
0bdfc961
NC
213=head2 compressed man pages
214
215Be able to install them. This would probably need a configure test to see how
216the system does compressed man pages (same directory/different directory?
217same filename/different filename), as well as tweaking the F<installman> script
218to compress as necessary.
219
30222c0f
NC
220=head2 Add a code coverage target to the Makefile
221
222Make it easy for anyone to run Devel::Cover on the core's tests. The steps
223to do this manually are roughly
224
225=over 4
226
227=item *
228
229do a normal C<Configure>, but include Devel::Cover as a module to install
230(see F<INSTALL> for how to do this)
231
232=item *
233
234 make perl
235
236=item *
237
238 cd t; HARNESS_PERL_SWITCHES=-MDevel::Cover ./perl -I../lib harness
239
240=item *
241
242Process the resulting Devel::Cover database
243
244=back
245
246This just give you the coverage of the F<.pm>s. To also get the C level
247coverage you need to
248
249=over 4
250
251=item *
252
253Additionally tell C<Configure> to use the appropriate C compiler flags for
254C<gcov>
255
256=item *
257
258 make perl.gcov
259
260(instead of C<make perl>)
261
262=item *
263
264After running the tests run C<gcov> to generate all the F<.gcov> files.
265(Including down in the subdirectories of F<ext/>
266
267=item *
268
269(From the top level perl directory) run C<gcov2perl> on all the C<.gcov> files
270to get their stats into the cover_db directory.
271
272=item *
273
274Then process the Devel::Cover database
275
276=back
277
278It would be good to add a single switch to C<Configure> to specify that you
279wanted to perform perl level coverage, and another to specify C level
280coverage, and have C<Configure> and the F<Makefile> do all the right things
281automatically.
282
02f21748 283=head2 Make Config.pm cope with differences between built and installed perl
0bdfc961
NC
284
285Quite often vendors ship a perl binary compiled with their (pay-for)
286compilers. People install a free compiler, such as gcc. To work out how to
287build extensions, Perl interrogates C<%Config>, so in this situation
288C<%Config> describes compilers that aren't there, and extension building
289fails. This forces people into choosing between re-compiling perl themselves
290using the compiler they have, or only using modules that the vendor ships.
291
292It would be good to find a way teach C<Config.pm> about the installation setup,
293possibly involving probing at install time or later, so that the C<%Config> in
294a binary distribution better describes the installed machine, when the
295installed machine differs from the build machine in some significant way.
296
728f4ecd
NC
297=head2 linker specification files
298
299Some platforms mandate that you provide a list of a shared library's external
300symbols to the linker, so the core already has the infrastructure in place to
301do this for generating shared perl libraries. My understanding is that the
302GNU toolchain can accept an optional linker specification file, and restrict
303visibility just to symbols declared in that file. It would be good to extend
304F<makedef.pl> to support this format, and to provide a means within
305C<Configure> to enable it. This would allow Unix users to test that the
306export list is correct, and to build a perl that does not pollute the global
307namespace with private symbols.
308
a229ae3b
RGS
309=head2 Cross-compile support
310
311Currently C<Configure> understands C<-Dusecrosscompile> option. This option
312arranges for building C<miniperl> for TARGET machine, so this C<miniperl> is
313assumed then to be copied to TARGET machine and used as a replacement of full
314C<perl> executable.
315
d1307786 316This could be done little differently. Namely C<miniperl> should be built for
a229ae3b 317HOST and then full C<perl> with extensions should be compiled for TARGET.
d1307786 318This, however, might require extra trickery for %Config: we have one config
87a942b1
JH
319first for HOST and then another for TARGET. Tools like MakeMaker will be
320mightily confused. Having around two different types of executables and
321libraries (HOST and TARGET) makes life interesting for Makefiles and
322shell (and Perl) scripts. There is $Config{run}, normally empty, which
323can be used as an execution wrapper. Also note that in some
324cross-compilation/execution environments the HOST and the TARGET do
325not see the same filesystem(s), the $Config{run} may need to do some
326file/directory copying back and forth.
0bdfc961 327
8537f021
RGS
328=head2 roffitall
329
330Make F<pod/roffitall> be updated by F<pod/buildtoc>.
331
98fca0e8
NC
332=head2 Split "linker" from "compiler"
333
334Right now, Configure probes for two commands, and sets two variables:
335
336=over 4
337
338=item * C<cc (cc.U)>
339
340This variable holds the name of a command to execute a C compiler which
341can resolve multiple global references that happen to have the same
342name. Usual values are F<cc> and F<gcc>.
343Fervent ANSI compilers may be called F<c89>. AIX has F<xlc>.
344
345=item * ld (dlsrc.U)
346
347This variable indicates the program to be used to link
348libraries for dynamic loading. On some systems, it is F<ld>.
349On ELF systems, it should be C<$cc>. Mostly, we'll try to respect
350the hint file setting.
351
352=back
353
354There is an implicit historical assumption, probably from Perl 1, that C<$cc>
355is also the correct command for linking object files together to make an
356executable. This may be true on Unix, but it's not true on other platforms,
357and there are a maze of work arounds in other places (such as F<Makefile.SH>)
358to cope with this.
359
360Ideally, we should create a new variable to hold the name of the executable
361linker program, probe for it in F<Configure>, and centralise all the special
362case logic there or in hints files.
363
364A small bikeshed issue remains - what to call it, given that C<$ld> is already
365taken (arguably for the wrong thing) and C<$link> could be confused with the
366Unix command line executable of the same name, which does something completely
367different. Andy Dougherty makes the counter argument "In parrot, I tried to
368call the command used to link object files and libraries into an executable
369F<link>, since that's what my vaguely-remembered DOS and VMS experience
370suggested. I don't think any real confusion has ensued, so it's probably a
371reasonable name for perl5 to use."
372
373"Alas, I've always worried that introducing it would make things worse,
374since now the module building utilities would have to look for
375C<$Config{link}> and institute a fall-back plan if it weren't found."
376
377
0bdfc961
NC
378=head1 Tasks that need a little C knowledge
379
380These tasks would need a little C knowledge, but don't need any specific
381background or experience with XS, or how the Perl interpreter works
382
3d826b29
NC
383=head2 Weed out needless PERL_UNUSED_ARG
384
385The C code uses the macro C<PERL_UNUSED_ARG> to stop compilers warning about
386unused arguments. Often the arguments can't be removed, as there is an
387external constraint that determines the prototype of the function, so this
388approach is valid. However, there are some cases where C<PERL_UNUSED_ARG>
389could be removed. Specifically
390
391=over 4
392
393=item *
394
395The prototypes of (nearly all) static functions can be changed
396
397=item *
398
399Unused arguments generated by short cut macros are wasteful - the short cut
400macro used can be changed.
401
402=back
403
fbf638cb
RGS
404=head2 Modernize the order of directories in @INC
405
406The way @INC is laid out by default, one cannot upgrade core (dual-life)
407modules without overwriting files. This causes problems for binary
3d14fd97
AD
408package builders. One possible proposal is laid out in this
409message:
410L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2002-04/msg02380.html>.
fbf638cb 411
bcbaa2d5
RGS
412=head2 -Duse32bit*
413
414Natively 64-bit systems need neither -Duse64bitint nor -Duse64bitall.
415On these systems, it might be the default compilation mode, and there
416is currently no guarantee that passing no use64bitall option to the
417Configure process will build a 32bit perl. Implementing -Duse32bit*
418options would be nice for perl 5.12.
419
0bdfc961 420=head2 Make it clear from -v if this is the exact official release
89007cb3 421
617eabfa
NC
422Currently perl from C<p4>/C<rsync> ships with a F<patchlevel.h> file that
423usually defines one local patch, of the form "MAINT12345" or "RC1". The output
424of perl -v doesn't report that a perl isn't an official release, and this
89007cb3 425information can get lost in bugs reports. Because of this, the minor version
fa11829f 426isn't bumped up until RC time, to minimise the possibility of versions of perl
89007cb3
NC
427escaping that believe themselves to be newer than they actually are.
428
429It would be useful to find an elegant way to have the "this is an interim
430maintenance release" or "this is a release candidate" in the terse -v output,
431and have it so that it's easy for the pumpking to remove this just as the
432release tarball is rolled up. This way the version pulled out of rsync would
433always say "I'm a development release" and it would be safe to bump the
434reported minor version as soon as a release ships, which would aid perl
435developers.
436
0bdfc961
NC
437This task is really about thinking of an elegant way to arrange the C source
438such that it's trivial for the Pumpking to flag "this is an official release"
439when making a tarball, yet leave the default source saying "I'm not the
440official release".
441
fee0a0f7 442=head2 Profile Perl - am I hot or not?
62403a3c 443
fee0a0f7
NC
444The Perl source code is stable enough that it makes sense to profile it,
445identify and optimise the hotspots. It would be good to measure the
446performance of the Perl interpreter using free tools such as cachegrind,
447gprof, and dtrace, and work to reduce the bottlenecks they reveal.
448
449As part of this, the idea of F<pp_hot.c> is that it contains the I<hot> ops,
450the ops that are most commonly used. The idea is that by grouping them, their
451object code will be adjacent in the executable, so they have a greater chance
452of already being in the CPU cache (or swapped in) due to being near another op
453already in use.
62403a3c
NC
454
455Except that it's not clear if these really are the most commonly used ops. So
fee0a0f7
NC
456as part of exercising your skills with coverage and profiling tools you might
457want to determine what ops I<really> are the most commonly used. And in turn
458suggest evictions and promotions to achieve a better F<pp_hot.c>.
62403a3c 459
91d0cbf6
NC
460One piece of Perl code that might make a good testbed is F<installman>.
461
98fed0ad
NC
462=head2 Allocate OPs from arenas
463
464Currently all new OP structures are individually malloc()ed and free()d.
465All C<malloc> implementations have space overheads, and are now as fast as
466custom allocates so it would both use less memory and less CPU to allocate
467the various OP structures from arenas. The SV arena code can probably be
468re-used for this.
469
539f2c54
JC
470Note that Configuring perl with C<-Accflags=-DPL_OP_SLAB_ALLOC> will use
471Perl_Slab_alloc() to pack optrees into a contiguous block, which is
472probably superior to the use of OP arenas, esp. from a cache locality
473standpoint. See L<Profile Perl - am I hot or not?>.
474
a229ae3b 475=head2 Improve win32/wince.c
0bdfc961 476
a229ae3b 477Currently, numerous functions look virtually, if not completely,
02f21748 478identical in both C<win32/wince.c> and C<win32/win32.c> files, which can't
6d71adcd
NC
479be good.
480
c5b31784
SH
481=head2 Use secure CRT functions when building with VC8 on Win32
482
483Visual C++ 2005 (VC++ 8.x) deprecated a number of CRT functions on the basis
484that they were "unsafe" and introduced differently named secure versions of
485them as replacements, e.g. instead of writing
486
487 FILE* f = fopen(__FILE__, "r");
488
489one should now write
490
491 FILE* f;
492 errno_t err = fopen_s(&f, __FILE__, "r");
493
494Currently, the warnings about these deprecations have been disabled by adding
495-D_CRT_SECURE_NO_DEPRECATE to the CFLAGS. It would be nice to remove that
496warning suppressant and actually make use of the new secure CRT functions.
497
498There is also a similar issue with POSIX CRT function names like fileno having
499been deprecated in favour of ISO C++ conformant names like _fileno. These
26a6faa8 500warnings are also currently suppressed by adding -D_CRT_NONSTDC_NO_DEPRECATE. It
c5b31784
SH
501might be nice to do as Microsoft suggest here too, although, unlike the secure
502functions issue, there is presumably little or no benefit in this case.
503
038ae9a4
SH
504=head2 Fix POSIX::access() and chdir() on Win32
505
506These functions currently take no account of DACLs and therefore do not behave
507correctly in situations where access is restricted by DACLs (as opposed to the
508read-only attribute).
509
510Furthermore, POSIX::access() behaves differently for directories having the
511read-only attribute set depending on what CRT library is being used. For
512example, the _access() function in the VC6 and VC7 CRTs (wrongly) claim that
513such directories are not writable, whereas in fact all directories are writable
514unless access is denied by DACLs. (In the case of directories, the read-only
515attribute actually only means that the directory cannot be deleted.) This CRT
516bug is fixed in the VC8 and VC9 CRTs (but, of course, the directory may still
517not actually be writable if access is indeed denied by DACLs).
518
519For the chdir() issue, see ActiveState bug #74552:
520http://bugs.activestate.com/show_bug.cgi?id=74552
521
522Therefore, DACLs should be checked both for consistency across CRTs and for
523the correct answer.
524
525(Note that perl's -w operator should not be modified to check DACLs. It has
526been written so that it reflects the state of the read-only attribute, even
527for directories (whatever CRT is being used), for symmetry with chmod().)
528
16815324
NC
529=head2 strcat(), strcpy(), strncat(), strncpy(), sprintf(), vsprintf()
530
531Maybe create a utility that checks after each libperl.a creation that
532none of the above (nor sprintf(), vsprintf(), or *SHUDDER* gets())
533ever creep back to libperl.a.
534
535 nm libperl.a | ./miniperl -alne '$o = $F[0] if /:$/; print "$o $F[1]" if $F[0] eq "U" && $F[1] =~ /^(?:strn?c(?:at|py)|v?sprintf|gets)$/'
536
537Note, of course, that this will only tell whether B<your> platform
538is using those naughty interfaces.
539
de96509d
JH
540=head2 -D_FORTIFY_SOURCE=2, -fstack-protector
541
542Recent glibcs support C<-D_FORTIFY_SOURCE=2> and recent gcc
543(4.1 onwards?) supports C<-fstack-protector>, both of which give
544protection against various kinds of buffer overflow problems.
545These should probably be used for compiling Perl whenever available,
546Configure and/or hints files should be adjusted to probe for the
547availability of these features and enable them as appropriate.
16815324 548
8964cfe0
NC
549=head2 Arenas for GPs? For MAGIC?
550
551C<struct gp> and C<struct magic> are both currently allocated by C<malloc>.
552It might be a speed or memory saving to change to using arenas. Or it might
553not. It would need some suitable benchmarking first. In particular, C<GP>s
554can probably be changed with minimal compatibility impact (probably nothing
555outside of the core, or even outside of F<gv.c> allocates them), but they
556probably aren't allocated/deallocated often enough for a speed saving. Whereas
557C<MAGIC> is allocated/deallocated more often, but in turn, is also something
558more externally visible, so changing the rules here may bite external code.
559
3880c8ec
NC
560=head2 Shared arenas
561
562Several SV body structs are now the same size, notably PVMG and PVGV, PVAV and
563PVHV, and PVCV and PVFM. It should be possible to allocate and return same
564sized bodies from the same actual arena, rather than maintaining one arena for
565each. This could save 4-6K per thread, of memory no longer tied up in the
566not-yet-allocated part of an arena.
567
8964cfe0 568
6d71adcd
NC
569=head1 Tasks that need a knowledge of XS
570
571These tasks would need C knowledge, and roughly the level of knowledge of
572the perl API that comes from writing modules that use XS to interface to
573C.
574
5d96f598
NC
575=head2 safely supporting POSIX SA_SIGINFO
576
577Some years ago Jarkko supplied patches to provide support for the POSIX
578SA_SIGINFO feature in Perl, passing the extra data to the Perl signal handler.
579
580Unfortunately, it only works with "unsafe" signals, because under safe
581signals, by the time Perl gets to run the signal handler, the extra
582information has been lost. Moreover, it's not easy to store it somewhere,
583as you can't call mutexs, or do anything else fancy, from inside a signal
584handler.
585
586So it strikes me that we could provide safe SA_SIGINFO support
587
588=over 4
589
590=item 1
591
592Provide global variables for two file descriptors
593
594=item 2
595
596When the first request is made via C<sigaction> for C<SA_SIGINFO>, create a
597pipe, store the reader in one, the writer in the other
598
599=item 3
600
601In the "safe" signal handler (C<Perl_csighandler()>/C<S_raise_signal()>), if
602the C<siginfo_t> pointer non-C<NULL>, and the writer file handle is open,
603
604=over 8
605
606=item 1
607
608serialise signal number, C<struct siginfo_t> (or at least the parts we care
609about) into a small auto char buff
610
611=item 2
612
613C<write()> that (non-blocking) to the writer fd
614
615=over 12
616
617=item 1
618
619if it writes 100%, flag the signal in a counter of "signals on the pipe" akin
620to the current per-signal-number counts
621
622=item 2
623
624if it writes 0%, assume the pipe is full. Flag the data as lost?
625
626=item 3
627
628if it writes partially, croak a panic, as your OS is broken.
629
630=back
631
632=back
633
634=item 4
635
636in the regular C<PERL_ASYNC_CHECK()> processing, if there are "signals on
637the pipe", read the data out, deserialise, build the Perl structures on
638the stack (code in C<Perl_sighandler()>, the "unsafe" handler), and call as
639usual.
640
641=back
642
643I think that this gets us decent C<SA_SIGINFO> support, without the current risk
644of running Perl code inside the signal handler context. (With all the dangers
645of things like C<malloc> corruption that that currently offers us)
646
647For more information see the thread starting with this message:
648http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-03/msg00305.html
649
6d71adcd
NC
650=head2 autovivification
651
652Make all autovivification consistent w.r.t LVALUE/RVALUE and strict/no strict;
653
654This task is incremental - even a little bit of work on it will help.
655
656=head2 Unicode in Filenames
657
658chdir, chmod, chown, chroot, exec, glob, link, lstat, mkdir, open,
659opendir, qx, readdir, readlink, rename, rmdir, stat, symlink, sysopen,
660system, truncate, unlink, utime, -X. All these could potentially accept
661Unicode filenames either as input or output (and in the case of system
662and qx Unicode in general, as input or output to/from the shell).
663Whether a filesystem - an operating system pair understands Unicode in
664filenames varies.
665
666Known combinations that have some level of understanding include
667Microsoft NTFS, Apple HFS+ (In Mac OS 9 and X) and Apple UFS (in Mac
668OS X), NFS v4 is rumored to be Unicode, and of course Plan 9. How to
669create Unicode filenames, what forms of Unicode are accepted and used
670(UCS-2, UTF-16, UTF-8), what (if any) is the normalization form used,
671and so on, varies. Finding the right level of interfacing to Perl
672requires some thought. Remember that an OS does not implicate a
673filesystem.
674
675(The Windows -C command flag "wide API support" has been at least
676temporarily retired in 5.8.1, and the -C has been repurposed, see
677L<perlrun>.)
678
87a942b1
JH
679Most probably the right way to do this would be this:
680L</"Virtualize operating system access">.
681
6d71adcd
NC
682=head2 Unicode in %ENV
683
684Currently the %ENV entries are always byte strings.
87a942b1 685See L</"Virtualize operating system access">.
6d71adcd 686
1f2e7916
JD
687=head2 Unicode and glob()
688
689Currently glob patterns and filenames returned from File::Glob::glob()
87a942b1 690are always byte strings. See L</"Virtualize operating system access">.
1f2e7916 691
dbb0c492
RGS
692=head2 Unicode and lc/uc operators
693
694Some built-in operators (C<lc>, C<uc>, etc.) behave differently, based on
695what the internal encoding of their argument is. That should not be the
696case. Maybe add a pragma to switch behaviour.
697
6d71adcd
NC
698=head2 use less 'memory'
699
700Investigate trade offs to switch out perl's choices on memory usage.
701Particularly perl should be able to give memory back.
702
703This task is incremental - even a little bit of work on it will help.
704
705=head2 Re-implement C<:unique> in a way that is actually thread-safe
706
707The old implementation made bad assumptions on several levels. A good 90%
708solution might be just to make C<:unique> work to share the string buffer
709of SvPVs. That way large constant strings can be shared between ithreads,
710such as the configuration information in F<Config>.
711
712=head2 Make tainting consistent
713
714Tainting would be easier to use if it didn't take documented shortcuts and
715allow taint to "leak" everywhere within an expression.
716
717=head2 readpipe(LIST)
718
719system() accepts a LIST syntax (and a PROGRAM LIST syntax) to avoid
720running a shell. readpipe() (the function behind qx//) could be similarly
721extended.
722
6d71adcd
NC
723=head2 Audit the code for destruction ordering assumptions
724
725Change 25773 notes
726
727 /* Need to check SvMAGICAL, as during global destruction it may be that
728 AvARYLEN(av) has been freed before av, and hence the SvANY() pointer
729 is now part of the linked list of SV heads, rather than pointing to
730 the original body. */
731 /* FIXME - audit the code for other bugs like this one. */
732
733adding the C<SvMAGICAL> check to
734
735 if (AvARYLEN(av) && SvMAGICAL(AvARYLEN(av))) {
736 MAGIC *mg = mg_find (AvARYLEN(av), PERL_MAGIC_arylen);
737
738Go through the core and look for similar assumptions that SVs have particular
739types, as all bets are off during global destruction.
740
749904bf
JH
741=head2 Extend PerlIO and PerlIO::Scalar
742
743PerlIO::Scalar doesn't know how to truncate(). Implementing this
744would require extending the PerlIO vtable.
745
746Similarly the PerlIO vtable doesn't know about formats (write()), or
747about stat(), or chmod()/chown(), utime(), or flock().
748
749(For PerlIO::Scalar it's hard to see what e.g. mode bits or ownership
750would mean.)
751
752PerlIO doesn't do directories or symlinks, either: mkdir(), rmdir(),
753opendir(), closedir(), seekdir(), rewinddir(), glob(); symlink(),
754readlink().
755
94da6c29
JH
756See also L</"Virtualize operating system access">.
757
3236f110
NC
758=head2 -C on the #! line
759
760It should be possible to make -C work correctly if found on the #! line,
761given that all perl command line options are strict ASCII, and -C changes
762only the interpretation of non-ASCII characters, and not for the script file
763handle. To make it work needs some investigation of the ordering of function
764calls during startup, and (by implication) a bit of tweaking of that order.
765
d6c1e11f
JH
766=head2 Organize error messages
767
768Perl's diagnostics (error messages, see L<perldiag>) could use
a8d0aeb9 769reorganizing and formalizing so that each error message has its
d6c1e11f
JH
770stable-for-all-eternity unique id, categorized by severity, type, and
771subsystem. (The error messages would be listed in a datafile outside
c4bd451b
CB
772of the Perl source code, and the source code would only refer to the
773messages by the id.) This clean-up and regularizing should apply
d6c1e11f
JH
774for all croak() messages.
775
776This would enable all sorts of things: easier translation/localization
777of the messages (though please do keep in mind the caveats of
778L<Locale::Maketext> about too straightforward approaches to
779translation), filtering by severity, and instead of grepping for a
780particular error message one could look for a stable error id. (Of
781course, changing the error messages by default would break all the
782existing software depending on some particular error message...)
783
784This kind of functionality is known as I<message catalogs>. Look for
785inspiration for example in the catgets() system, possibly even use it
786if available-- but B<only> if available, all platforms will B<not>
de96509d 787have catgets().
d6c1e11f
JH
788
789For the really pure at heart, consider extending this item to cover
790also the warning messages (see L<perllexwarn>, C<warnings.pl>).
3236f110 791
0bdfc961 792=head1 Tasks that need a knowledge of the interpreter
3298bd4d 793
0bdfc961
NC
794These tasks would need C knowledge, and knowledge of how the interpreter works,
795or a willingness to learn.
3298bd4d 796
718140ec
NC
797=head2 lexicals used only once
798
799This warns:
800
801 $ perl -we '$pie = 42'
802 Name "main::pie" used only once: possible typo at -e line 1.
803
804This does not:
805
806 $ perl -we 'my $pie = 42'
807
808Logically all lexicals used only once should warn, if the user asks for
d6f4ea2e
SP
809warnings. An unworked RT ticket (#5087) has been open for almost seven
810years for this discrepancy.
718140ec 811
a3d15f9a
RGS
812=head2 UTF-8 revamp
813
814The handling of Unicode is unclean in many places. For example, the regexp
815engine matches in Unicode semantics whenever the string or the pattern is
816flagged as UTF-8, but that should not be dependent on an internal storage
817detail of the string. Likewise, case folding behaviour is dependent on the
818UTF8 internal flag being on or off.
819
820=head2 Properly Unicode safe tokeniser and pads.
821
822The tokeniser isn't actually very UTF-8 clean. C<use utf8;> is a hack -
823variable names are stored in stashes as raw bytes, without the utf-8 flag
824set. The pad API only takes a C<char *> pointer, so that's all bytes too. The
825tokeniser ignores the UTF-8-ness of C<PL_rsfp>, or any SVs returned from
826source filters. All this could be fixed.
827
636e63cb
NC
828=head2 state variable initialization in list context
829
830Currently this is illegal:
831
832 state ($a, $b) = foo();
833
a2874905 834In Perl 6, C<state ($a) = foo();> and C<(state $a) = foo();> have different
a8d0aeb9 835semantics, which is tricky to implement in Perl 5 as currently they produce
a2874905 836the same opcode trees. The Perl 6 design is firm, so it would be good to
a8d0aeb9 837implement the necessary code in Perl 5. There are comments in
a2874905
NC
838C<Perl_newASSIGNOP()> that show the code paths taken by various assignment
839constructions involving state variables.
636e63cb 840
4fedb12c
RGS
841=head2 Implement $value ~~ 0 .. $range
842
843It would be nice to extend the syntax of the C<~~> operator to also
844understand numeric (and maybe alphanumeric) ranges.
a393eb28
RGS
845
846=head2 A does() built-in
847
848Like ref(), only useful. It would call the C<DOES> method on objects; it
849would also tell whether something can be dereferenced as an
850array/hash/etc., or used as a regexp, etc.
851L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-03/msg00481.html>
852
853=head2 Tied filehandles and write() don't mix
854
855There is no method on tied filehandles to allow them to be called back by
856formats.
4fedb12c 857
d10fc472 858=head2 Attach/detach debugger from running program
1626a787 859
cd793d32
NC
860The old perltodo notes "With C<gdb>, you can attach the debugger to a running
861program if you pass the process ID. It would be good to do this with the Perl
0bdfc961
NC
862debugger on a running Perl program, although I'm not sure how it would be
863done." ssh and screen do this with named pipes in /tmp. Maybe we can too.
1626a787 864
a8cb5b9e
RGS
865=head2 Optimize away empty destructors
866
867Defining an empty DESTROY method might be useful (notably in
868AUTOLOAD-enabled classes), but it's still a bit expensive to call. That
869could probably be optimized.
870
0bdfc961
NC
871=head2 LVALUE functions for lists
872
873The old perltodo notes that lvalue functions don't work for list or hash
874slices. This would be good to fix.
875
876=head2 LVALUE functions in the debugger
877
878The old perltodo notes that lvalue functions don't work in the debugger. This
879would be good to fix.
880
0bdfc961
NC
881=head2 regexp optimiser optional
882
883The regexp optimiser is not optional. It should configurable to be, to allow
884its performance to be measured, and its bugs to be easily demonstrated.
885
02f21748
RGS
886=head2 delete &function
887
888Allow to delete functions. One can already undef them, but they're still
889in the stash.
890
ef36c6a7
RGS
891=head2 C</w> regex modifier
892
893That flag would enable to match whole words, and also to interpolate
894arrays as alternations. With it, C</P/w> would be roughly equivalent to:
895
896 do { local $"='|'; /\b(?:P)\b/ }
897
898See L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-01/msg00400.html>
899for the discussion.
900
0bdfc961
NC
901=head2 optional optimizer
902
903Make the peephole optimizer optional. Currently it performs two tasks as
904it walks the optree - genuine peephole optimisations, and necessary fixups of
905ops. It would be good to find an efficient way to switch out the
906optimisations whilst keeping the fixups.
907
908=head2 You WANT *how* many
909
910Currently contexts are void, scalar and list. split has a special mechanism in
911place to pass in the number of return values wanted. It would be useful to
912have a general mechanism for this, backwards compatible and little speed hit.
913This would allow proposals such as short circuiting sort to be implemented
914as a module on CPAN.
915
916=head2 lexical aliases
917
918Allow lexical aliases (maybe via the syntax C<my \$alias = \$foo>.
919
920=head2 entersub XS vs Perl
921
922At the moment pp_entersub is huge, and has code to deal with entering both
923perl and XS subroutines. Subroutine implementations rarely change between
924perl and XS at run time, so investigate using 2 ops to enter subs (one for
925XS, one for perl) and swap between if a sub is redefined.
2810d901 926
de535794 927=head2 Self-ties
2810d901 928
de535794 929Self-ties are currently illegal because they caused too many segfaults. Maybe
a8d0aeb9 930the causes of these could be tracked down and self-ties on all types
de535794 931reinstated.
0bdfc961
NC
932
933=head2 Optimize away @_
934
935The old perltodo notes "Look at the "reification" code in C<av.c>".
936
87a942b1
JH
937=head2 Virtualize operating system access
938
939Implement a set of "vtables" that virtualizes operating system access
940(open(), mkdir(), unlink(), readdir(), getenv(), etc.) At the very
941least these interfaces should take SVs as "name" arguments instead of
942bare char pointers; probably the most flexible and extensible way
e1a3d5d1
JH
943would be for the Perl-facing interfaces to accept HVs. The system
944needs to be per-operating-system and per-file-system
945hookable/filterable, preferably both from XS and Perl level
87a942b1
JH
946(L<perlport/"Files and Filesystems"> is good reading at this point,
947in fact, all of L<perlport> is.)
948
e1a3d5d1
JH
949This has actually already been implemented (but only for Win32),
950take a look at F<iperlsys.h> and F<win32/perlhost.h>. While all Win32
951variants go through a set of "vtables" for operating system access,
952non-Win32 systems currently go straight for the POSIX/UNIX-style
953system/library call. Similar system as for Win32 should be
954implemented for all platforms. The existing Win32 implementation
955probably does not need to survive alongside this proposed new
956implementation, the approaches could be merged.
87a942b1
JH
957
958What would this give us? One often-asked-for feature this would
94da6c29
JH
959enable is using Unicode for filenames, and other "names" like %ENV,
960usernames, hostnames, and so forth.
961(See L<perlunicode/"When Unicode Does Not Happen">.)
962
963But this kind of virtualization would also allow for things like
964virtual filesystems, virtual networks, and "sandboxes" (though as long
965as dynamic loading of random object code is allowed, not very safe
966sandboxes since external code of course know not of Perl's vtables).
967An example of a smaller "sandbox" is that this feature can be used to
968implement per-thread working directories: Win32 already does this.
969
970See also L</"Extend PerlIO and PerlIO::Scalar">.
87a942b1 971
ac6197af
NC
972=head2 Investigate PADTMP hash pessimisation
973
974The peephole optimier converts constants used for hash key lookups to shared
057163d7 975hash key scalars. Under ithreads, something is undoing this work.
ac6197af
NC
976See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-09/msg00793.html
977
057163d7
NC
978=head2 Store the current pad in the OP slab allocator
979
980=for clarification
981I hope that I got that "current pad" part correct
982
983Currently we leak ops in various cases of parse failure. I suggested that we
984could solve this by always using the op slab allocator, and walking it to
985free ops. Dave comments that as some ops are already freed during optree
986creation one would have to mark which ops are freed, and not double free them
987when walking the slab. He notes that one problem with this is that for some ops
988you have to know which pad was current at the time of allocation, which does
989change. I suggested storing a pointer to the current pad in the memory allocated
990for the slab, and swapping to a new slab each time the pad changes. Dave thinks
991that this would work.
992
52960e22
JC
993=head2 repack the optree
994
995Repacking the optree after execution order is determined could allow
057163d7
NC
996removal of NULL ops, and optimal ordering of OPs with respect to cache-line
997filling. The slab allocator could be reused for this purpose. I think that
998the best way to do this is to make it an optional step just before the
999completed optree is attached to anything else, and to use the slab allocator
1000unchanged, so that freeing ops is identical whether or not this step runs.
1001Note that the slab allocator allocates ops downwards in memory, so one would
1002have to actually "allocate" the ops in reverse-execution order to get them
1003contiguous in memory in execution order.
1004
1005See http://www.nntp.perl.org/group/perl.perl5.porters/2007/12/msg131975.html
1006
1007Note that running this copy, and then freeing all the old location ops would
1008cause their slabs to be freed, which would eliminate possible memory wastage if
1009the previous suggestion is implemented, and we swap slabs more frequently.
52960e22 1010
12e06b6f
NC
1011=head2 eliminate incorrect line numbers in warnings
1012
1013This code
1014
1015 use warnings;
1016 my $undef;
1017
1018 if ($undef == 3) {
1019 } elsif ($undef == 0) {
1020 }
1021
18a16cc5 1022used to produce this output:
12e06b6f
NC
1023
1024 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1025 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1026
18a16cc5
NC
1027where the line of the second warning was misreported - it should be line 5.
1028Rafael fixed this - the problem arose because there was no nextstate OP
1029between the execution of the C<if> and the C<elsif>, hence C<PL_curcop> still
1030reports that the currently executing line is line 4. The solution was to inject
1031a nextstate OPs for each C<elsif>, although it turned out that the nextstate
1032OP needed to be a nulled OP, rather than a live nextstate OP, else other line
1033numbers became misreported. (Jenga!)
12e06b6f
NC
1034
1035The problem is more general than C<elsif> (although the C<elsif> case is the
1036most common and the most confusing). Ideally this code
1037
1038 use warnings;
1039 my $undef;
1040
1041 my $a = $undef + 1;
1042 my $b
1043 = $undef
1044 + 1;
1045
1046would produce this output
1047
1048 Use of uninitialized value $undef in addition (+) at wrong.pl line 4.
1049 Use of uninitialized value $undef in addition (+) at wrong.pl line 7.
1050
1051(rather than lines 4 and 5), but this would seem to require every OP to carry
1052(at least) line number information.
1053
1054What might work is to have an optional line number in memory just before the
1055BASEOP structure, with a flag bit in the op to say whether it's present.
1056Initially during compile every OP would carry its line number. Then add a late
1057pass to the optimiser (potentially combined with L</repack the optree>) which
1058looks at the two ops on every edge of the graph of the execution path. If
1059the line number changes, flags the destination OP with this information.
1060Once all paths are traced, replace every op with the flag with a
1061nextstate-light op (that just updates C<PL_curcop>), which in turn then passes
1062control on to the true op. All ops would then be replaced by variants that
1063do not store the line number. (Which, logically, why it would work best in
1064conjunction with L</repack the optree>, as that is already copying/reallocating
1065all the OPs)
1066
18a16cc5
NC
1067(Although I should note that we're not certain that doing this for the general
1068case is worth it)
1069
52960e22
JC
1070=head2 optimize tail-calls
1071
1072Tail-calls present an opportunity for broadly applicable optimization;
1073anywhere that C<< return foo(...) >> is called, the outer return can
1074be replaced by a goto, and foo will return directly to the outer
1075caller, saving (conservatively) 25% of perl's call&return cost, which
1076is relatively higher than in C. The scheme language is known to do
1077this heavily. B::Concise provides good insight into where this
1078optimization is possible, ie anywhere entersub,leavesub op-sequence
1079occurs.
1080
1081 perl -MO=Concise,-exec,a,b,-main -e 'sub a{ 1 }; sub b {a()}; b(2)'
1082
1083Bottom line on this is probably a new pp_tailcall function which
1084combines the code in pp_entersub, pp_leavesub. This should probably
1085be done 1st in XS, and using B::Generate to patch the new OP into the
1086optrees.
1087
0bdfc961
NC
1088=head1 Big projects
1089
1090Tasks that will get your name mentioned in the description of the "Highlights
87a942b1 1091of 5.12"
0bdfc961
NC
1092
1093=head2 make ithreads more robust
1094
4e577f8b 1095Generally make ithreads more robust. See also L</iCOW>
0bdfc961
NC
1096
1097This task is incremental - even a little bit of work on it will help, and
1098will be greatly appreciated.
1099
6c047da7
YST
1100One bit would be to write the missing code in sv.c:Perl_dirp_dup.
1101
59c7f7d5
RGS
1102Fix Perl_sv_dup, et al so that threads can return objects.
1103
0bdfc961
NC
1104=head2 iCOW
1105
1106Sarathy and Arthur have a proposal for an improved Copy On Write which
1107specifically will be able to COW new ithreads. If this can be implemented
1108it would be a good thing.
1109
1110=head2 (?{...}) closures in regexps
1111
1112Fix (or rewrite) the implementation of the C</(?{...})/> closures.
1113
1114=head2 A re-entrant regexp engine
1115
1116This will allow the use of a regex from inside (?{ }), (??{ }) and
1117(?(?{ })|) constructs.
6bda09f9 1118
6bda09f9
YO
1119=head2 Add class set operations to regexp engine
1120
1121Apparently these are quite useful. Anyway, Jeffery Friedl wants them.
1122
1123demerphq has this on his todo list, but right at the bottom.