This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Tweaks and reflowing for the cc/linker split.
[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
162f8c67
NC
26=head2 Smartmatch design issues
27
28In 5.10.0 the smartmatch operator C<~~> isn't working quite "right". But
29before we can fix the implementation, we need to define what "right" is.
30The first problem is that Robin Houston implemented the Perl 6 smart match
31spec as of February 2006, when smart match was axiomatically symmetrical:
32L<http://groups.google.com/group/perl.perl6.language/msg/bf2b486f089ad021>
33
34Since then the Perl 6 target moved, but the Perl 5 implementation did not.
35
36So it would be useful for someone to compare the Perl 6 smartmatch table
37as of February 2006 L<http://svn.perl.org/viewvc/perl6/doc/trunk/design/syn/S03.pod?view=markup&pathrev=7615>
38and the current table L<http://svn.perl.org/viewvc/perl6/doc/trunk/design/syn/S03.pod?revision=14556&view=markup>
39and tabulate the differences in Perl 6. The diff is
40C<svn diff -r7615:14556 http://svn.perl.org/perl6/doc/trunk/design/syn/S03.pod>
41-- search for C<=head1 Smart matching>
42
43With that done and published, someone (else) can then map any changed Perl 6
44semantics back to Perl 5, based on how the existing semantics map to Perl 5:
45L<http://search.cpan.org/~rgarcia/perl-5.10.0/pod/perlsyn.pod#Smart_matching_in_detail>
46
47
48There are also some questions that need answering:
49
50=over 4
51
52=item *
53
54How do you negate one? (documentation issue)
55http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-01/msg00071.html
56
57=item *
58
59Array behaviors
60http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-12/msg00799.html
61
62* Should smart matches be symmetrical? (Perl 6 says no)
63
64* Other differences between Perl 5 and Perl 6 smart match?
65
66=item *
67
68Objects and smart match
69http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-12/msg00865.html
70
71=back
72
5a176cbc
NC
73=head2 Remove duplication of test setup.
74
75Schwern notes, that there's duplication of code - lots and lots of tests have
76some variation on the big block of C<$Is_Foo> checks. We can safely put this
77into a file, change it to build an C<%Is> hash and require it. Maybe just put
78it into F<test.pl>. Throw in the handy tainting subroutines.
79
87a942b1 80=head2 POD -E<gt> HTML conversion in the core still sucks
e50bb9a1 81
938c8732 82Which is crazy given just how simple POD purports to be, and how simple HTML
adebf063
NC
83can be. It's not actually I<as> simple as it sounds, particularly with the
84flexibility POD allows for C<=item>, but it would be good to improve the
85visual appeal of the HTML generated, and to avoid it having any validation
86errors. See also L</make HTML install work>, as the layout of installation tree
87is needed to improve the cross-linking.
938c8732 88
dc0fb092
SP
89The addition of C<Pod::Simple> and its related modules may make this task
90easier to complete.
91
8537f021
RGS
92=head2 merge checkpods and podchecker
93
94F<pod/checkpods.PL> (and C<make check> in the F<pod/> subdirectory)
95implements a very basic check for pod files, but the errors it discovers
96aren't found by podchecker. Add this check to podchecker, get rid of
97checkpods and have C<make check> use podchecker.
98
aa237293
NC
99=head2 Parallel testing
100
b2e2905c 101(This probably impacts much more than the core: also the Test::Harness
02f21748
RGS
102and TAP::* modules on CPAN.)
103
c707cc00
NC
104All of the tests in F<t/> can now be run in parallel, if C<$ENV{TEST_JOBS}>
105is set. However, tests within each directory in F<ext> and F<lib> are still
106run in series, with directories run in parallel. This is an adequate
107heuristic, but it might be possible to relax it further, and get more
108throughput. Specifically, it would be good to audit all of F<lib/*.t>, and
109make them use C<File::Temp>.
aa237293 110
0bdfc961 111=head2 Make Schwern poorer
e50bb9a1 112
613bd4f7 113We should have tests for everything. When all the core's modules are tested,
0bdfc961
NC
114Schwern has promised to donate to $500 to TPF. We may need volunteers to
115hold him upside down and shake vigorously in order to actually extract the
116cash.
3958b146 117
0bdfc961 118=head2 Improve the coverage of the core tests
e50bb9a1 119
02f21748
RGS
120Use Devel::Cover to ascertain the core modules's test coverage, then add
121tests that are currently missing.
30222c0f 122
0bdfc961 123=head2 test B
e50bb9a1 124
0bdfc961 125A full test suite for the B module would be nice.
e50bb9a1 126
636e63cb
NC
127=head2 Deparse inlined constants
128
129Code such as this
130
131 use constant PI => 4;
132 warn PI
133
134will currently deparse as
135
136 use constant ('PI', 4);
137 warn 4;
138
139because the tokenizer inlines the value of the constant subroutine C<PI>.
140This allows various compile time optimisations, such as constant folding
141and dead code elimination. Where these haven't happened (such as the example
142above) it ought be possible to make B::Deparse work out the name of the
143original constant, because just enough information survives in the symbol
144table to do this. Specifically, the same scalar is used for the constant in
145the optree as is used for the constant subroutine, so by iterating over all
146symbol tables and generating a mapping of SV address to constant name, it
147would be possible to provide B::Deparse with this functionality.
148
0bdfc961 149=head2 A decent benchmark
e50bb9a1 150
617eabfa 151C<perlbench> seems impervious to any recent changes made to the perl core. It
0bdfc961
NC
152would be useful to have a reasonable general benchmarking suite that roughly
153represented what current perl programs do, and measurably reported whether
154tweaks to the core improve, degrade or don't really affect performance, to
155guide people attempting to optimise the guts of perl. Gisle would welcome
156new tests for perlbench.
6168cf99 157
0bdfc961 158=head2 fix tainting bugs
6168cf99 159
0bdfc961
NC
160Fix the bugs revealed by running the test suite with the C<-t> switch (via
161C<make test.taintwarn>).
e50bb9a1 162
0bdfc961 163=head2 Dual life everything
e50bb9a1 164
0bdfc961
NC
165As part of the "dists" plan, anything that doesn't belong in the smallest perl
166distribution needs to be dual lifed. Anything else can be too. Figure out what
167changes would be needed to package that module and its tests up for CPAN, and
168do so. Test it with older perl releases, and fix the problems you find.
e50bb9a1 169
a393eb28
RGS
170To make a minimal perl distribution, it's useful to look at
171F<t/lib/commonsense.t>.
172
c2aba5b8
RGS
173=head2 Bundle dual life modules in ext/
174
175For maintenance (and branch merging) reasons, it would be useful to move
176some architecture-independent dual-life modules from lib/ to ext/, if this
177has no negative impact on the build of perl itself.
178
0bdfc961 179=head2 POSIX memory footprint
e50bb9a1 180
0bdfc961
NC
181Ilya observed that use POSIX; eats memory like there's no tomorrow, and at
182various times worked to cut it down. There is probably still fat to cut out -
183for example POSIX passes Exporter some very memory hungry data structures.
e50bb9a1 184
eed36644
NC
185=head2 embed.pl/makedef.pl
186
187There is a script F<embed.pl> that generates several header files to prefix
188all of Perl's symbols in a consistent way, to provide some semblance of
189namespace support in C<C>. Functions are declared in F<embed.fnc>, variables
907b3e23 190in F<interpvar.h>. Quite a few of the functions and variables
eed36644
NC
191are conditionally declared there, using C<#ifdef>. However, F<embed.pl>
192doesn't understand the C macros, so the rules about which symbols are present
193when is duplicated in F<makedef.pl>. Writing things twice is bad, m'kay.
194It would be good to teach C<embed.pl> to understand the conditional
195compilation, and hence remove the duplication, and the mistakes it has caused.
e50bb9a1 196
801de10e
NC
197=head2 use strict; and AutoLoad
198
199Currently if you write
200
201 package Whack;
202 use AutoLoader 'AUTOLOAD';
203 use strict;
204 1;
205 __END__
206 sub bloop {
207 print join (' ', No, strict, here), "!\n";
208 }
209
210then C<use strict;> isn't in force within the autoloaded subroutines. It would
211be more consistent (and less surprising) to arrange for all lexical pragmas
212in force at the __END__ block to be in force within each autoloaded subroutine.
213
773b3597
RGS
214There's a similar problem with SelfLoader.
215
91d0cbf6
NC
216=head2 profile installman
217
218The F<installman> script is slow. All it is doing text processing, which we're
219told is something Perl is good at. So it would be nice to know what it is doing
220that is taking so much CPU, and where possible address it.
221
222
0bdfc961 223=head1 Tasks that need a little sysadmin-type knowledge
e50bb9a1 224
0bdfc961
NC
225Or if you prefer, tasks that you would learn from, and broaden your skills
226base...
e50bb9a1 227
cd793d32 228=head2 make HTML install work
e50bb9a1 229
adebf063
NC
230There is an C<installhtml> target in the Makefile. It's marked as
231"experimental". It would be good to get this tested, make it work reliably, and
232remove the "experimental" tag. This would include
233
234=over 4
235
236=item 1
237
238Checking that cross linking between various parts of the documentation works.
239In particular that links work between the modules (files with POD in F<lib/>)
240and the core documentation (files in F<pod/>)
241
242=item 2
243
617eabfa
NC
244Work out how to split C<perlfunc> into chunks, preferably one per function
245group, preferably with general case code that could be used elsewhere.
246Challenges here are correctly identifying the groups of functions that go
247together, and making the right named external cross-links point to the right
248page. Things to be aware of are C<-X>, groups such as C<getpwnam> to
249C<endservent>, two or more C<=items> giving the different parameter lists, such
250as
adebf063
NC
251
252 =item substr EXPR,OFFSET,LENGTH,REPLACEMENT
adebf063 253 =item substr EXPR,OFFSET,LENGTH
adebf063
NC
254 =item substr EXPR,OFFSET
255
256and different parameter lists having different meanings. (eg C<select>)
257
258=back
3a89a73c 259
0bdfc961
NC
260=head2 compressed man pages
261
262Be able to install them. This would probably need a configure test to see how
263the system does compressed man pages (same directory/different directory?
264same filename/different filename), as well as tweaking the F<installman> script
265to compress as necessary.
266
30222c0f
NC
267=head2 Add a code coverage target to the Makefile
268
269Make it easy for anyone to run Devel::Cover on the core's tests. The steps
270to do this manually are roughly
271
272=over 4
273
274=item *
275
276do a normal C<Configure>, but include Devel::Cover as a module to install
277(see F<INSTALL> for how to do this)
278
279=item *
280
281 make perl
282
283=item *
284
285 cd t; HARNESS_PERL_SWITCHES=-MDevel::Cover ./perl -I../lib harness
286
287=item *
288
289Process the resulting Devel::Cover database
290
291=back
292
293This just give you the coverage of the F<.pm>s. To also get the C level
294coverage you need to
295
296=over 4
297
298=item *
299
300Additionally tell C<Configure> to use the appropriate C compiler flags for
301C<gcov>
302
303=item *
304
305 make perl.gcov
306
307(instead of C<make perl>)
308
309=item *
310
311After running the tests run C<gcov> to generate all the F<.gcov> files.
312(Including down in the subdirectories of F<ext/>
313
314=item *
315
316(From the top level perl directory) run C<gcov2perl> on all the C<.gcov> files
317to get their stats into the cover_db directory.
318
319=item *
320
321Then process the Devel::Cover database
322
323=back
324
325It would be good to add a single switch to C<Configure> to specify that you
326wanted to perform perl level coverage, and another to specify C level
327coverage, and have C<Configure> and the F<Makefile> do all the right things
328automatically.
329
02f21748 330=head2 Make Config.pm cope with differences between built and installed perl
0bdfc961
NC
331
332Quite often vendors ship a perl binary compiled with their (pay-for)
333compilers. People install a free compiler, such as gcc. To work out how to
334build extensions, Perl interrogates C<%Config>, so in this situation
335C<%Config> describes compilers that aren't there, and extension building
336fails. This forces people into choosing between re-compiling perl themselves
337using the compiler they have, or only using modules that the vendor ships.
338
339It would be good to find a way teach C<Config.pm> about the installation setup,
340possibly involving probing at install time or later, so that the C<%Config> in
341a binary distribution better describes the installed machine, when the
342installed machine differs from the build machine in some significant way.
343
728f4ecd
NC
344=head2 linker specification files
345
346Some platforms mandate that you provide a list of a shared library's external
347symbols to the linker, so the core already has the infrastructure in place to
348do this for generating shared perl libraries. My understanding is that the
349GNU toolchain can accept an optional linker specification file, and restrict
350visibility just to symbols declared in that file. It would be good to extend
351F<makedef.pl> to support this format, and to provide a means within
352C<Configure> to enable it. This would allow Unix users to test that the
353export list is correct, and to build a perl that does not pollute the global
354namespace with private symbols.
355
a229ae3b
RGS
356=head2 Cross-compile support
357
358Currently C<Configure> understands C<-Dusecrosscompile> option. This option
359arranges for building C<miniperl> for TARGET machine, so this C<miniperl> is
360assumed then to be copied to TARGET machine and used as a replacement of full
361C<perl> executable.
362
d1307786 363This could be done little differently. Namely C<miniperl> should be built for
a229ae3b 364HOST and then full C<perl> with extensions should be compiled for TARGET.
d1307786 365This, however, might require extra trickery for %Config: we have one config
87a942b1
JH
366first for HOST and then another for TARGET. Tools like MakeMaker will be
367mightily confused. Having around two different types of executables and
368libraries (HOST and TARGET) makes life interesting for Makefiles and
369shell (and Perl) scripts. There is $Config{run}, normally empty, which
370can be used as an execution wrapper. Also note that in some
371cross-compilation/execution environments the HOST and the TARGET do
372not see the same filesystem(s), the $Config{run} may need to do some
373file/directory copying back and forth.
0bdfc961 374
8537f021
RGS
375=head2 roffitall
376
377Make F<pod/roffitall> be updated by F<pod/buildtoc>.
378
98fca0e8
NC
379=head2 Split "linker" from "compiler"
380
381Right now, Configure probes for two commands, and sets two variables:
382
383=over 4
384
385=item * C<cc (cc.U)>
386
387This variable holds the name of a command to execute a C compiler which
388can resolve multiple global references that happen to have the same
389name. Usual values are F<cc> and F<gcc>.
390Fervent ANSI compilers may be called F<c89>. AIX has F<xlc>.
391
392=item * ld (dlsrc.U)
393
394This variable indicates the program to be used to link
395libraries for dynamic loading. On some systems, it is F<ld>.
396On ELF systems, it should be C<$cc>. Mostly, we'll try to respect
397the hint file setting.
398
399=back
400
8d159ec1
NC
401There is an implicit historical assumption from around Perl5.000alpha
402something, that C<$cc> is also the correct command for linking object files
403together to make an executable. This may be true on Unix, but it's not true
404on other platforms, and there are a maze of work arounds in other places (such
405as F<Makefile.SH>) to cope with this.
98fca0e8
NC
406
407Ideally, we should create a new variable to hold the name of the executable
408linker program, probe for it in F<Configure>, and centralise all the special
409case logic there or in hints files.
410
411A small bikeshed issue remains - what to call it, given that C<$ld> is already
8d159ec1
NC
412taken (arguably for the wrong thing now, but on SunOS 4.1 it is the command
413for creating dynamically-loadable modules) and C<$link> could be confused with
414the Unix command line executable of the same name, which does something
415completely different. Andy Dougherty makes the counter argument "In parrot, I
416tried to call the command used to link object files and libraries into an
417executable F<link>, since that's what my vaguely-remembered DOS and VMS
418experience suggested. I don't think any real confusion has ensued, so it's
419probably a reasonable name for perl5 to use."
98fca0e8
NC
420
421"Alas, I've always worried that introducing it would make things worse,
422since now the module building utilities would have to look for
423C<$Config{link}> and institute a fall-back plan if it weren't found."
8d159ec1
NC
424Although I can see that as confusing, given that C<$Config{d_link}> is true
425when (hard) links are available.
98fca0e8 426
0bdfc961
NC
427=head1 Tasks that need a little C knowledge
428
429These tasks would need a little C knowledge, but don't need any specific
430background or experience with XS, or how the Perl interpreter works
431
3d826b29
NC
432=head2 Weed out needless PERL_UNUSED_ARG
433
434The C code uses the macro C<PERL_UNUSED_ARG> to stop compilers warning about
435unused arguments. Often the arguments can't be removed, as there is an
436external constraint that determines the prototype of the function, so this
437approach is valid. However, there are some cases where C<PERL_UNUSED_ARG>
438could be removed. Specifically
439
440=over 4
441
442=item *
443
444The prototypes of (nearly all) static functions can be changed
445
446=item *
447
448Unused arguments generated by short cut macros are wasteful - the short cut
449macro used can be changed.
450
451=back
452
fbf638cb
RGS
453=head2 Modernize the order of directories in @INC
454
455The way @INC is laid out by default, one cannot upgrade core (dual-life)
456modules without overwriting files. This causes problems for binary
3d14fd97
AD
457package builders. One possible proposal is laid out in this
458message:
459L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2002-04/msg02380.html>.
fbf638cb 460
bcbaa2d5
RGS
461=head2 -Duse32bit*
462
463Natively 64-bit systems need neither -Duse64bitint nor -Duse64bitall.
464On these systems, it might be the default compilation mode, and there
465is currently no guarantee that passing no use64bitall option to the
466Configure process will build a 32bit perl. Implementing -Duse32bit*
467options would be nice for perl 5.12.
468
0bdfc961 469=head2 Make it clear from -v if this is the exact official release
89007cb3 470
617eabfa
NC
471Currently perl from C<p4>/C<rsync> ships with a F<patchlevel.h> file that
472usually defines one local patch, of the form "MAINT12345" or "RC1". The output
473of perl -v doesn't report that a perl isn't an official release, and this
89007cb3 474information can get lost in bugs reports. Because of this, the minor version
fa11829f 475isn't bumped up until RC time, to minimise the possibility of versions of perl
89007cb3
NC
476escaping that believe themselves to be newer than they actually are.
477
478It would be useful to find an elegant way to have the "this is an interim
479maintenance release" or "this is a release candidate" in the terse -v output,
480and have it so that it's easy for the pumpking to remove this just as the
481release tarball is rolled up. This way the version pulled out of rsync would
482always say "I'm a development release" and it would be safe to bump the
483reported minor version as soon as a release ships, which would aid perl
484developers.
485
0bdfc961
NC
486This task is really about thinking of an elegant way to arrange the C source
487such that it's trivial for the Pumpking to flag "this is an official release"
488when making a tarball, yet leave the default source saying "I'm not the
489official release".
490
fee0a0f7 491=head2 Profile Perl - am I hot or not?
62403a3c 492
fee0a0f7
NC
493The Perl source code is stable enough that it makes sense to profile it,
494identify and optimise the hotspots. It would be good to measure the
495performance of the Perl interpreter using free tools such as cachegrind,
496gprof, and dtrace, and work to reduce the bottlenecks they reveal.
497
498As part of this, the idea of F<pp_hot.c> is that it contains the I<hot> ops,
499the ops that are most commonly used. The idea is that by grouping them, their
500object code will be adjacent in the executable, so they have a greater chance
501of already being in the CPU cache (or swapped in) due to being near another op
502already in use.
62403a3c
NC
503
504Except that it's not clear if these really are the most commonly used ops. So
fee0a0f7
NC
505as part of exercising your skills with coverage and profiling tools you might
506want to determine what ops I<really> are the most commonly used. And in turn
507suggest evictions and promotions to achieve a better F<pp_hot.c>.
62403a3c 508
91d0cbf6
NC
509One piece of Perl code that might make a good testbed is F<installman>.
510
98fed0ad
NC
511=head2 Allocate OPs from arenas
512
513Currently all new OP structures are individually malloc()ed and free()d.
514All C<malloc> implementations have space overheads, and are now as fast as
515custom allocates so it would both use less memory and less CPU to allocate
516the various OP structures from arenas. The SV arena code can probably be
517re-used for this.
518
539f2c54
JC
519Note that Configuring perl with C<-Accflags=-DPL_OP_SLAB_ALLOC> will use
520Perl_Slab_alloc() to pack optrees into a contiguous block, which is
521probably superior to the use of OP arenas, esp. from a cache locality
522standpoint. See L<Profile Perl - am I hot or not?>.
523
a229ae3b 524=head2 Improve win32/wince.c
0bdfc961 525
a229ae3b 526Currently, numerous functions look virtually, if not completely,
02f21748 527identical in both C<win32/wince.c> and C<win32/win32.c> files, which can't
6d71adcd
NC
528be good.
529
c5b31784
SH
530=head2 Use secure CRT functions when building with VC8 on Win32
531
532Visual C++ 2005 (VC++ 8.x) deprecated a number of CRT functions on the basis
533that they were "unsafe" and introduced differently named secure versions of
534them as replacements, e.g. instead of writing
535
536 FILE* f = fopen(__FILE__, "r");
537
538one should now write
539
540 FILE* f;
541 errno_t err = fopen_s(&f, __FILE__, "r");
542
543Currently, the warnings about these deprecations have been disabled by adding
544-D_CRT_SECURE_NO_DEPRECATE to the CFLAGS. It would be nice to remove that
545warning suppressant and actually make use of the new secure CRT functions.
546
547There is also a similar issue with POSIX CRT function names like fileno having
548been deprecated in favour of ISO C++ conformant names like _fileno. These
26a6faa8 549warnings are also currently suppressed by adding -D_CRT_NONSTDC_NO_DEPRECATE. It
c5b31784
SH
550might be nice to do as Microsoft suggest here too, although, unlike the secure
551functions issue, there is presumably little or no benefit in this case.
552
038ae9a4
SH
553=head2 Fix POSIX::access() and chdir() on Win32
554
555These functions currently take no account of DACLs and therefore do not behave
556correctly in situations where access is restricted by DACLs (as opposed to the
557read-only attribute).
558
559Furthermore, POSIX::access() behaves differently for directories having the
560read-only attribute set depending on what CRT library is being used. For
561example, the _access() function in the VC6 and VC7 CRTs (wrongly) claim that
562such directories are not writable, whereas in fact all directories are writable
563unless access is denied by DACLs. (In the case of directories, the read-only
564attribute actually only means that the directory cannot be deleted.) This CRT
565bug is fixed in the VC8 and VC9 CRTs (but, of course, the directory may still
566not actually be writable if access is indeed denied by DACLs).
567
568For the chdir() issue, see ActiveState bug #74552:
569http://bugs.activestate.com/show_bug.cgi?id=74552
570
571Therefore, DACLs should be checked both for consistency across CRTs and for
572the correct answer.
573
574(Note that perl's -w operator should not be modified to check DACLs. It has
575been written so that it reflects the state of the read-only attribute, even
576for directories (whatever CRT is being used), for symmetry with chmod().)
577
16815324
NC
578=head2 strcat(), strcpy(), strncat(), strncpy(), sprintf(), vsprintf()
579
580Maybe create a utility that checks after each libperl.a creation that
581none of the above (nor sprintf(), vsprintf(), or *SHUDDER* gets())
582ever creep back to libperl.a.
583
584 nm libperl.a | ./miniperl -alne '$o = $F[0] if /:$/; print "$o $F[1]" if $F[0] eq "U" && $F[1] =~ /^(?:strn?c(?:at|py)|v?sprintf|gets)$/'
585
586Note, of course, that this will only tell whether B<your> platform
587is using those naughty interfaces.
588
de96509d
JH
589=head2 -D_FORTIFY_SOURCE=2, -fstack-protector
590
591Recent glibcs support C<-D_FORTIFY_SOURCE=2> and recent gcc
592(4.1 onwards?) supports C<-fstack-protector>, both of which give
593protection against various kinds of buffer overflow problems.
594These should probably be used for compiling Perl whenever available,
595Configure and/or hints files should be adjusted to probe for the
596availability of these features and enable them as appropriate.
16815324 597
8964cfe0
NC
598=head2 Arenas for GPs? For MAGIC?
599
600C<struct gp> and C<struct magic> are both currently allocated by C<malloc>.
601It might be a speed or memory saving to change to using arenas. Or it might
602not. It would need some suitable benchmarking first. In particular, C<GP>s
603can probably be changed with minimal compatibility impact (probably nothing
604outside of the core, or even outside of F<gv.c> allocates them), but they
605probably aren't allocated/deallocated often enough for a speed saving. Whereas
606C<MAGIC> is allocated/deallocated more often, but in turn, is also something
607more externally visible, so changing the rules here may bite external code.
608
3880c8ec
NC
609=head2 Shared arenas
610
611Several SV body structs are now the same size, notably PVMG and PVGV, PVAV and
612PVHV, and PVCV and PVFM. It should be possible to allocate and return same
613sized bodies from the same actual arena, rather than maintaining one arena for
614each. This could save 4-6K per thread, of memory no longer tied up in the
615not-yet-allocated part of an arena.
616
8964cfe0 617
6d71adcd
NC
618=head1 Tasks that need a knowledge of XS
619
620These tasks would need C knowledge, and roughly the level of knowledge of
621the perl API that comes from writing modules that use XS to interface to
622C.
623
5d96f598
NC
624=head2 safely supporting POSIX SA_SIGINFO
625
626Some years ago Jarkko supplied patches to provide support for the POSIX
627SA_SIGINFO feature in Perl, passing the extra data to the Perl signal handler.
628
629Unfortunately, it only works with "unsafe" signals, because under safe
630signals, by the time Perl gets to run the signal handler, the extra
631information has been lost. Moreover, it's not easy to store it somewhere,
632as you can't call mutexs, or do anything else fancy, from inside a signal
633handler.
634
635So it strikes me that we could provide safe SA_SIGINFO support
636
637=over 4
638
639=item 1
640
641Provide global variables for two file descriptors
642
643=item 2
644
645When the first request is made via C<sigaction> for C<SA_SIGINFO>, create a
646pipe, store the reader in one, the writer in the other
647
648=item 3
649
650In the "safe" signal handler (C<Perl_csighandler()>/C<S_raise_signal()>), if
651the C<siginfo_t> pointer non-C<NULL>, and the writer file handle is open,
652
653=over 8
654
655=item 1
656
657serialise signal number, C<struct siginfo_t> (or at least the parts we care
658about) into a small auto char buff
659
660=item 2
661
662C<write()> that (non-blocking) to the writer fd
663
664=over 12
665
666=item 1
667
668if it writes 100%, flag the signal in a counter of "signals on the pipe" akin
669to the current per-signal-number counts
670
671=item 2
672
673if it writes 0%, assume the pipe is full. Flag the data as lost?
674
675=item 3
676
677if it writes partially, croak a panic, as your OS is broken.
678
679=back
680
681=back
682
683=item 4
684
685in the regular C<PERL_ASYNC_CHECK()> processing, if there are "signals on
686the pipe", read the data out, deserialise, build the Perl structures on
687the stack (code in C<Perl_sighandler()>, the "unsafe" handler), and call as
688usual.
689
690=back
691
692I think that this gets us decent C<SA_SIGINFO> support, without the current risk
693of running Perl code inside the signal handler context. (With all the dangers
694of things like C<malloc> corruption that that currently offers us)
695
696For more information see the thread starting with this message:
697http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-03/msg00305.html
698
6d71adcd
NC
699=head2 autovivification
700
701Make all autovivification consistent w.r.t LVALUE/RVALUE and strict/no strict;
702
703This task is incremental - even a little bit of work on it will help.
704
705=head2 Unicode in Filenames
706
707chdir, chmod, chown, chroot, exec, glob, link, lstat, mkdir, open,
708opendir, qx, readdir, readlink, rename, rmdir, stat, symlink, sysopen,
709system, truncate, unlink, utime, -X. All these could potentially accept
710Unicode filenames either as input or output (and in the case of system
711and qx Unicode in general, as input or output to/from the shell).
712Whether a filesystem - an operating system pair understands Unicode in
713filenames varies.
714
715Known combinations that have some level of understanding include
716Microsoft NTFS, Apple HFS+ (In Mac OS 9 and X) and Apple UFS (in Mac
717OS X), NFS v4 is rumored to be Unicode, and of course Plan 9. How to
718create Unicode filenames, what forms of Unicode are accepted and used
719(UCS-2, UTF-16, UTF-8), what (if any) is the normalization form used,
720and so on, varies. Finding the right level of interfacing to Perl
721requires some thought. Remember that an OS does not implicate a
722filesystem.
723
724(The Windows -C command flag "wide API support" has been at least
725temporarily retired in 5.8.1, and the -C has been repurposed, see
726L<perlrun>.)
727
87a942b1
JH
728Most probably the right way to do this would be this:
729L</"Virtualize operating system access">.
730
6d71adcd
NC
731=head2 Unicode in %ENV
732
733Currently the %ENV entries are always byte strings.
87a942b1 734See L</"Virtualize operating system access">.
6d71adcd 735
1f2e7916
JD
736=head2 Unicode and glob()
737
738Currently glob patterns and filenames returned from File::Glob::glob()
87a942b1 739are always byte strings. See L</"Virtualize operating system access">.
1f2e7916 740
dbb0c492
RGS
741=head2 Unicode and lc/uc operators
742
743Some built-in operators (C<lc>, C<uc>, etc.) behave differently, based on
744what the internal encoding of their argument is. That should not be the
745case. Maybe add a pragma to switch behaviour.
746
6d71adcd
NC
747=head2 use less 'memory'
748
749Investigate trade offs to switch out perl's choices on memory usage.
750Particularly perl should be able to give memory back.
751
752This task is incremental - even a little bit of work on it will help.
753
754=head2 Re-implement C<:unique> in a way that is actually thread-safe
755
756The old implementation made bad assumptions on several levels. A good 90%
757solution might be just to make C<:unique> work to share the string buffer
758of SvPVs. That way large constant strings can be shared between ithreads,
759such as the configuration information in F<Config>.
760
761=head2 Make tainting consistent
762
763Tainting would be easier to use if it didn't take documented shortcuts and
764allow taint to "leak" everywhere within an expression.
765
766=head2 readpipe(LIST)
767
768system() accepts a LIST syntax (and a PROGRAM LIST syntax) to avoid
769running a shell. readpipe() (the function behind qx//) could be similarly
770extended.
771
6d71adcd
NC
772=head2 Audit the code for destruction ordering assumptions
773
774Change 25773 notes
775
776 /* Need to check SvMAGICAL, as during global destruction it may be that
777 AvARYLEN(av) has been freed before av, and hence the SvANY() pointer
778 is now part of the linked list of SV heads, rather than pointing to
779 the original body. */
780 /* FIXME - audit the code for other bugs like this one. */
781
782adding the C<SvMAGICAL> check to
783
784 if (AvARYLEN(av) && SvMAGICAL(AvARYLEN(av))) {
785 MAGIC *mg = mg_find (AvARYLEN(av), PERL_MAGIC_arylen);
786
787Go through the core and look for similar assumptions that SVs have particular
788types, as all bets are off during global destruction.
789
749904bf
JH
790=head2 Extend PerlIO and PerlIO::Scalar
791
792PerlIO::Scalar doesn't know how to truncate(). Implementing this
793would require extending the PerlIO vtable.
794
795Similarly the PerlIO vtable doesn't know about formats (write()), or
796about stat(), or chmod()/chown(), utime(), or flock().
797
798(For PerlIO::Scalar it's hard to see what e.g. mode bits or ownership
799would mean.)
800
801PerlIO doesn't do directories or symlinks, either: mkdir(), rmdir(),
802opendir(), closedir(), seekdir(), rewinddir(), glob(); symlink(),
803readlink().
804
94da6c29
JH
805See also L</"Virtualize operating system access">.
806
3236f110
NC
807=head2 -C on the #! line
808
809It should be possible to make -C work correctly if found on the #! line,
810given that all perl command line options are strict ASCII, and -C changes
811only the interpretation of non-ASCII characters, and not for the script file
812handle. To make it work needs some investigation of the ordering of function
813calls during startup, and (by implication) a bit of tweaking of that order.
814
d6c1e11f
JH
815=head2 Organize error messages
816
817Perl's diagnostics (error messages, see L<perldiag>) could use
a8d0aeb9 818reorganizing and formalizing so that each error message has its
d6c1e11f
JH
819stable-for-all-eternity unique id, categorized by severity, type, and
820subsystem. (The error messages would be listed in a datafile outside
c4bd451b
CB
821of the Perl source code, and the source code would only refer to the
822messages by the id.) This clean-up and regularizing should apply
d6c1e11f
JH
823for all croak() messages.
824
825This would enable all sorts of things: easier translation/localization
826of the messages (though please do keep in mind the caveats of
827L<Locale::Maketext> about too straightforward approaches to
828translation), filtering by severity, and instead of grepping for a
829particular error message one could look for a stable error id. (Of
830course, changing the error messages by default would break all the
831existing software depending on some particular error message...)
832
833This kind of functionality is known as I<message catalogs>. Look for
834inspiration for example in the catgets() system, possibly even use it
835if available-- but B<only> if available, all platforms will B<not>
de96509d 836have catgets().
d6c1e11f
JH
837
838For the really pure at heart, consider extending this item to cover
839also the warning messages (see L<perllexwarn>, C<warnings.pl>).
3236f110 840
0bdfc961 841=head1 Tasks that need a knowledge of the interpreter
3298bd4d 842
0bdfc961
NC
843These tasks would need C knowledge, and knowledge of how the interpreter works,
844or a willingness to learn.
3298bd4d 845
718140ec
NC
846=head2 lexicals used only once
847
848This warns:
849
850 $ perl -we '$pie = 42'
851 Name "main::pie" used only once: possible typo at -e line 1.
852
853This does not:
854
855 $ perl -we 'my $pie = 42'
856
857Logically all lexicals used only once should warn, if the user asks for
d6f4ea2e
SP
858warnings. An unworked RT ticket (#5087) has been open for almost seven
859years for this discrepancy.
718140ec 860
a3d15f9a
RGS
861=head2 UTF-8 revamp
862
863The handling of Unicode is unclean in many places. For example, the regexp
864engine matches in Unicode semantics whenever the string or the pattern is
865flagged as UTF-8, but that should not be dependent on an internal storage
866detail of the string. Likewise, case folding behaviour is dependent on the
867UTF8 internal flag being on or off.
868
869=head2 Properly Unicode safe tokeniser and pads.
870
871The tokeniser isn't actually very UTF-8 clean. C<use utf8;> is a hack -
872variable names are stored in stashes as raw bytes, without the utf-8 flag
873set. The pad API only takes a C<char *> pointer, so that's all bytes too. The
874tokeniser ignores the UTF-8-ness of C<PL_rsfp>, or any SVs returned from
875source filters. All this could be fixed.
876
636e63cb
NC
877=head2 state variable initialization in list context
878
879Currently this is illegal:
880
881 state ($a, $b) = foo();
882
a2874905 883In Perl 6, C<state ($a) = foo();> and C<(state $a) = foo();> have different
a8d0aeb9 884semantics, which is tricky to implement in Perl 5 as currently they produce
a2874905 885the same opcode trees. The Perl 6 design is firm, so it would be good to
a8d0aeb9 886implement the necessary code in Perl 5. There are comments in
a2874905
NC
887C<Perl_newASSIGNOP()> that show the code paths taken by various assignment
888constructions involving state variables.
636e63cb 889
4fedb12c
RGS
890=head2 Implement $value ~~ 0 .. $range
891
892It would be nice to extend the syntax of the C<~~> operator to also
893understand numeric (and maybe alphanumeric) ranges.
a393eb28
RGS
894
895=head2 A does() built-in
896
897Like ref(), only useful. It would call the C<DOES> method on objects; it
898would also tell whether something can be dereferenced as an
899array/hash/etc., or used as a regexp, etc.
900L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-03/msg00481.html>
901
902=head2 Tied filehandles and write() don't mix
903
904There is no method on tied filehandles to allow them to be called back by
905formats.
4fedb12c 906
d10fc472 907=head2 Attach/detach debugger from running program
1626a787 908
cd793d32
NC
909The old perltodo notes "With C<gdb>, you can attach the debugger to a running
910program if you pass the process ID. It would be good to do this with the Perl
0bdfc961
NC
911debugger on a running Perl program, although I'm not sure how it would be
912done." ssh and screen do this with named pipes in /tmp. Maybe we can too.
1626a787 913
a8cb5b9e
RGS
914=head2 Optimize away empty destructors
915
916Defining an empty DESTROY method might be useful (notably in
917AUTOLOAD-enabled classes), but it's still a bit expensive to call. That
918could probably be optimized.
919
0bdfc961
NC
920=head2 LVALUE functions for lists
921
922The old perltodo notes that lvalue functions don't work for list or hash
923slices. This would be good to fix.
924
925=head2 LVALUE functions in the debugger
926
927The old perltodo notes that lvalue functions don't work in the debugger. This
928would be good to fix.
929
0bdfc961
NC
930=head2 regexp optimiser optional
931
932The regexp optimiser is not optional. It should configurable to be, to allow
933its performance to be measured, and its bugs to be easily demonstrated.
934
02f21748
RGS
935=head2 delete &function
936
937Allow to delete functions. One can already undef them, but they're still
938in the stash.
939
ef36c6a7
RGS
940=head2 C</w> regex modifier
941
942That flag would enable to match whole words, and also to interpolate
943arrays as alternations. With it, C</P/w> would be roughly equivalent to:
944
945 do { local $"='|'; /\b(?:P)\b/ }
946
947See L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-01/msg00400.html>
948for the discussion.
949
0bdfc961
NC
950=head2 optional optimizer
951
952Make the peephole optimizer optional. Currently it performs two tasks as
953it walks the optree - genuine peephole optimisations, and necessary fixups of
954ops. It would be good to find an efficient way to switch out the
955optimisations whilst keeping the fixups.
956
957=head2 You WANT *how* many
958
959Currently contexts are void, scalar and list. split has a special mechanism in
960place to pass in the number of return values wanted. It would be useful to
961have a general mechanism for this, backwards compatible and little speed hit.
962This would allow proposals such as short circuiting sort to be implemented
963as a module on CPAN.
964
965=head2 lexical aliases
966
967Allow lexical aliases (maybe via the syntax C<my \$alias = \$foo>.
968
969=head2 entersub XS vs Perl
970
971At the moment pp_entersub is huge, and has code to deal with entering both
972perl and XS subroutines. Subroutine implementations rarely change between
973perl and XS at run time, so investigate using 2 ops to enter subs (one for
974XS, one for perl) and swap between if a sub is redefined.
2810d901 975
de535794 976=head2 Self-ties
2810d901 977
de535794 978Self-ties are currently illegal because they caused too many segfaults. Maybe
a8d0aeb9 979the causes of these could be tracked down and self-ties on all types
de535794 980reinstated.
0bdfc961
NC
981
982=head2 Optimize away @_
983
984The old perltodo notes "Look at the "reification" code in C<av.c>".
985
87a942b1
JH
986=head2 Virtualize operating system access
987
988Implement a set of "vtables" that virtualizes operating system access
989(open(), mkdir(), unlink(), readdir(), getenv(), etc.) At the very
990least these interfaces should take SVs as "name" arguments instead of
991bare char pointers; probably the most flexible and extensible way
e1a3d5d1
JH
992would be for the Perl-facing interfaces to accept HVs. The system
993needs to be per-operating-system and per-file-system
994hookable/filterable, preferably both from XS and Perl level
87a942b1
JH
995(L<perlport/"Files and Filesystems"> is good reading at this point,
996in fact, all of L<perlport> is.)
997
e1a3d5d1
JH
998This has actually already been implemented (but only for Win32),
999take a look at F<iperlsys.h> and F<win32/perlhost.h>. While all Win32
1000variants go through a set of "vtables" for operating system access,
1001non-Win32 systems currently go straight for the POSIX/UNIX-style
1002system/library call. Similar system as for Win32 should be
1003implemented for all platforms. The existing Win32 implementation
1004probably does not need to survive alongside this proposed new
1005implementation, the approaches could be merged.
87a942b1
JH
1006
1007What would this give us? One often-asked-for feature this would
94da6c29
JH
1008enable is using Unicode for filenames, and other "names" like %ENV,
1009usernames, hostnames, and so forth.
1010(See L<perlunicode/"When Unicode Does Not Happen">.)
1011
1012But this kind of virtualization would also allow for things like
1013virtual filesystems, virtual networks, and "sandboxes" (though as long
1014as dynamic loading of random object code is allowed, not very safe
1015sandboxes since external code of course know not of Perl's vtables).
1016An example of a smaller "sandbox" is that this feature can be used to
1017implement per-thread working directories: Win32 already does this.
1018
1019See also L</"Extend PerlIO and PerlIO::Scalar">.
87a942b1 1020
ac6197af
NC
1021=head2 Investigate PADTMP hash pessimisation
1022
1023The peephole optimier converts constants used for hash key lookups to shared
057163d7 1024hash key scalars. Under ithreads, something is undoing this work.
ac6197af
NC
1025See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-09/msg00793.html
1026
057163d7
NC
1027=head2 Store the current pad in the OP slab allocator
1028
1029=for clarification
1030I hope that I got that "current pad" part correct
1031
1032Currently we leak ops in various cases of parse failure. I suggested that we
1033could solve this by always using the op slab allocator, and walking it to
1034free ops. Dave comments that as some ops are already freed during optree
1035creation one would have to mark which ops are freed, and not double free them
1036when walking the slab. He notes that one problem with this is that for some ops
1037you have to know which pad was current at the time of allocation, which does
1038change. I suggested storing a pointer to the current pad in the memory allocated
1039for the slab, and swapping to a new slab each time the pad changes. Dave thinks
1040that this would work.
1041
52960e22
JC
1042=head2 repack the optree
1043
1044Repacking the optree after execution order is determined could allow
057163d7
NC
1045removal of NULL ops, and optimal ordering of OPs with respect to cache-line
1046filling. The slab allocator could be reused for this purpose. I think that
1047the best way to do this is to make it an optional step just before the
1048completed optree is attached to anything else, and to use the slab allocator
1049unchanged, so that freeing ops is identical whether or not this step runs.
1050Note that the slab allocator allocates ops downwards in memory, so one would
1051have to actually "allocate" the ops in reverse-execution order to get them
1052contiguous in memory in execution order.
1053
1054See http://www.nntp.perl.org/group/perl.perl5.porters/2007/12/msg131975.html
1055
1056Note that running this copy, and then freeing all the old location ops would
1057cause their slabs to be freed, which would eliminate possible memory wastage if
1058the previous suggestion is implemented, and we swap slabs more frequently.
52960e22 1059
12e06b6f
NC
1060=head2 eliminate incorrect line numbers in warnings
1061
1062This code
1063
1064 use warnings;
1065 my $undef;
1066
1067 if ($undef == 3) {
1068 } elsif ($undef == 0) {
1069 }
1070
18a16cc5 1071used to produce this output:
12e06b6f
NC
1072
1073 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1074 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1075
18a16cc5
NC
1076where the line of the second warning was misreported - it should be line 5.
1077Rafael fixed this - the problem arose because there was no nextstate OP
1078between the execution of the C<if> and the C<elsif>, hence C<PL_curcop> still
1079reports that the currently executing line is line 4. The solution was to inject
1080a nextstate OPs for each C<elsif>, although it turned out that the nextstate
1081OP needed to be a nulled OP, rather than a live nextstate OP, else other line
1082numbers became misreported. (Jenga!)
12e06b6f
NC
1083
1084The problem is more general than C<elsif> (although the C<elsif> case is the
1085most common and the most confusing). Ideally this code
1086
1087 use warnings;
1088 my $undef;
1089
1090 my $a = $undef + 1;
1091 my $b
1092 = $undef
1093 + 1;
1094
1095would produce this output
1096
1097 Use of uninitialized value $undef in addition (+) at wrong.pl line 4.
1098 Use of uninitialized value $undef in addition (+) at wrong.pl line 7.
1099
1100(rather than lines 4 and 5), but this would seem to require every OP to carry
1101(at least) line number information.
1102
1103What might work is to have an optional line number in memory just before the
1104BASEOP structure, with a flag bit in the op to say whether it's present.
1105Initially during compile every OP would carry its line number. Then add a late
1106pass to the optimiser (potentially combined with L</repack the optree>) which
1107looks at the two ops on every edge of the graph of the execution path. If
1108the line number changes, flags the destination OP with this information.
1109Once all paths are traced, replace every op with the flag with a
1110nextstate-light op (that just updates C<PL_curcop>), which in turn then passes
1111control on to the true op. All ops would then be replaced by variants that
1112do not store the line number. (Which, logically, why it would work best in
1113conjunction with L</repack the optree>, as that is already copying/reallocating
1114all the OPs)
1115
18a16cc5
NC
1116(Although I should note that we're not certain that doing this for the general
1117case is worth it)
1118
52960e22
JC
1119=head2 optimize tail-calls
1120
1121Tail-calls present an opportunity for broadly applicable optimization;
1122anywhere that C<< return foo(...) >> is called, the outer return can
1123be replaced by a goto, and foo will return directly to the outer
1124caller, saving (conservatively) 25% of perl's call&return cost, which
1125is relatively higher than in C. The scheme language is known to do
1126this heavily. B::Concise provides good insight into where this
1127optimization is possible, ie anywhere entersub,leavesub op-sequence
1128occurs.
1129
1130 perl -MO=Concise,-exec,a,b,-main -e 'sub a{ 1 }; sub b {a()}; b(2)'
1131
1132Bottom line on this is probably a new pp_tailcall function which
1133combines the code in pp_entersub, pp_leavesub. This should probably
1134be done 1st in XS, and using B::Generate to patch the new OP into the
1135optrees.
1136
0bdfc961
NC
1137=head1 Big projects
1138
1139Tasks that will get your name mentioned in the description of the "Highlights
87a942b1 1140of 5.12"
0bdfc961
NC
1141
1142=head2 make ithreads more robust
1143
4e577f8b 1144Generally make ithreads more robust. See also L</iCOW>
0bdfc961
NC
1145
1146This task is incremental - even a little bit of work on it will help, and
1147will be greatly appreciated.
1148
6c047da7
YST
1149One bit would be to write the missing code in sv.c:Perl_dirp_dup.
1150
59c7f7d5
RGS
1151Fix Perl_sv_dup, et al so that threads can return objects.
1152
0bdfc961
NC
1153=head2 iCOW
1154
1155Sarathy and Arthur have a proposal for an improved Copy On Write which
1156specifically will be able to COW new ithreads. If this can be implemented
1157it would be a good thing.
1158
1159=head2 (?{...}) closures in regexps
1160
1161Fix (or rewrite) the implementation of the C</(?{...})/> closures.
1162
1163=head2 A re-entrant regexp engine
1164
1165This will allow the use of a regex from inside (?{ }), (??{ }) and
1166(?(?{ })|) constructs.
6bda09f9 1167
6bda09f9
YO
1168=head2 Add class set operations to regexp engine
1169
1170Apparently these are quite useful. Anyway, Jeffery Friedl wants them.
1171
1172demerphq has this on his todo list, but right at the bottom.