This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Note the possibility of sharing arenas between types.
[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
aa237293
NC
57The core regression test suite is getting ever more comprehensive, which has
58the side effect that it takes longer to run. This isn't so good. Investigate
59whether it would be feasible to give the harness script the B<option> of
60running sets of tests in parallel. This would be useful for tests in
61F<t/op/*.t> and F<t/uni/*.t> and maybe some sets of tests in F<lib/>.
62
63Questions to answer
64
65=over 4
66
67=item 1
68
69How does screen layout work when you're running more than one test?
70
71=item 2
72
73How does the caller of test specify how many tests to run in parallel?
74
75=item 3
76
77How do setup/teardown tests identify themselves?
78
79=back
80
81Pugs already does parallel testing - can their approach be re-used?
82
0bdfc961 83=head2 Make Schwern poorer
e50bb9a1 84
613bd4f7 85We should have tests for everything. When all the core's modules are tested,
0bdfc961
NC
86Schwern has promised to donate to $500 to TPF. We may need volunteers to
87hold him upside down and shake vigorously in order to actually extract the
88cash.
3958b146 89
0bdfc961 90=head2 Improve the coverage of the core tests
e50bb9a1 91
02f21748
RGS
92Use Devel::Cover to ascertain the core modules's test coverage, then add
93tests that are currently missing.
30222c0f 94
0bdfc961 95=head2 test B
e50bb9a1 96
0bdfc961 97A full test suite for the B module would be nice.
e50bb9a1 98
636e63cb
NC
99=head2 Deparse inlined constants
100
101Code such as this
102
103 use constant PI => 4;
104 warn PI
105
106will currently deparse as
107
108 use constant ('PI', 4);
109 warn 4;
110
111because the tokenizer inlines the value of the constant subroutine C<PI>.
112This allows various compile time optimisations, such as constant folding
113and dead code elimination. Where these haven't happened (such as the example
114above) it ought be possible to make B::Deparse work out the name of the
115original constant, because just enough information survives in the symbol
116table to do this. Specifically, the same scalar is used for the constant in
117the optree as is used for the constant subroutine, so by iterating over all
118symbol tables and generating a mapping of SV address to constant name, it
119would be possible to provide B::Deparse with this functionality.
120
0bdfc961 121=head2 A decent benchmark
e50bb9a1 122
617eabfa 123C<perlbench> seems impervious to any recent changes made to the perl core. It
0bdfc961
NC
124would be useful to have a reasonable general benchmarking suite that roughly
125represented what current perl programs do, and measurably reported whether
126tweaks to the core improve, degrade or don't really affect performance, to
127guide people attempting to optimise the guts of perl. Gisle would welcome
128new tests for perlbench.
6168cf99 129
0bdfc961 130=head2 fix tainting bugs
6168cf99 131
0bdfc961
NC
132Fix the bugs revealed by running the test suite with the C<-t> switch (via
133C<make test.taintwarn>).
e50bb9a1 134
0bdfc961 135=head2 Dual life everything
e50bb9a1 136
0bdfc961
NC
137As part of the "dists" plan, anything that doesn't belong in the smallest perl
138distribution needs to be dual lifed. Anything else can be too. Figure out what
139changes would be needed to package that module and its tests up for CPAN, and
140do so. Test it with older perl releases, and fix the problems you find.
e50bb9a1 141
a393eb28
RGS
142To make a minimal perl distribution, it's useful to look at
143F<t/lib/commonsense.t>.
144
c2aba5b8
RGS
145=head2 Bundle dual life modules in ext/
146
147For maintenance (and branch merging) reasons, it would be useful to move
148some architecture-independent dual-life modules from lib/ to ext/, if this
149has no negative impact on the build of perl itself.
150
151However, we need to make sure that they are still installed in
152architecture-independent directories by C<make install>.
153
0bdfc961 154=head2 Improving C<threads::shared>
722d2a37 155
0bdfc961
NC
156Investigate whether C<threads::shared> could share aggregates properly with
157only Perl level changes to shared.pm
722d2a37 158
0bdfc961 159=head2 POSIX memory footprint
e50bb9a1 160
0bdfc961
NC
161Ilya observed that use POSIX; eats memory like there's no tomorrow, and at
162various times worked to cut it down. There is probably still fat to cut out -
163for example POSIX passes Exporter some very memory hungry data structures.
e50bb9a1 164
eed36644
NC
165=head2 embed.pl/makedef.pl
166
167There is a script F<embed.pl> that generates several header files to prefix
168all of Perl's symbols in a consistent way, to provide some semblance of
169namespace support in C<C>. Functions are declared in F<embed.fnc>, variables
907b3e23 170in F<interpvar.h>. Quite a few of the functions and variables
eed36644
NC
171are conditionally declared there, using C<#ifdef>. However, F<embed.pl>
172doesn't understand the C macros, so the rules about which symbols are present
173when is duplicated in F<makedef.pl>. Writing things twice is bad, m'kay.
174It would be good to teach C<embed.pl> to understand the conditional
175compilation, and hence remove the duplication, and the mistakes it has caused.
e50bb9a1 176
801de10e
NC
177=head2 use strict; and AutoLoad
178
179Currently if you write
180
181 package Whack;
182 use AutoLoader 'AUTOLOAD';
183 use strict;
184 1;
185 __END__
186 sub bloop {
187 print join (' ', No, strict, here), "!\n";
188 }
189
190then C<use strict;> isn't in force within the autoloaded subroutines. It would
191be more consistent (and less surprising) to arrange for all lexical pragmas
192in force at the __END__ block to be in force within each autoloaded subroutine.
193
773b3597
RGS
194There's a similar problem with SelfLoader.
195
91d0cbf6
NC
196=head2 profile installman
197
198The F<installman> script is slow. All it is doing text processing, which we're
199told is something Perl is good at. So it would be nice to know what it is doing
200that is taking so much CPU, and where possible address it.
201
202
0bdfc961 203=head1 Tasks that need a little sysadmin-type knowledge
e50bb9a1 204
0bdfc961
NC
205Or if you prefer, tasks that you would learn from, and broaden your skills
206base...
e50bb9a1 207
cd793d32 208=head2 make HTML install work
e50bb9a1 209
adebf063
NC
210There is an C<installhtml> target in the Makefile. It's marked as
211"experimental". It would be good to get this tested, make it work reliably, and
212remove the "experimental" tag. This would include
213
214=over 4
215
216=item 1
217
218Checking that cross linking between various parts of the documentation works.
219In particular that links work between the modules (files with POD in F<lib/>)
220and the core documentation (files in F<pod/>)
221
222=item 2
223
617eabfa
NC
224Work out how to split C<perlfunc> into chunks, preferably one per function
225group, preferably with general case code that could be used elsewhere.
226Challenges here are correctly identifying the groups of functions that go
227together, and making the right named external cross-links point to the right
228page. Things to be aware of are C<-X>, groups such as C<getpwnam> to
229C<endservent>, two or more C<=items> giving the different parameter lists, such
230as
adebf063
NC
231
232 =item substr EXPR,OFFSET,LENGTH,REPLACEMENT
adebf063 233 =item substr EXPR,OFFSET,LENGTH
adebf063
NC
234 =item substr EXPR,OFFSET
235
236and different parameter lists having different meanings. (eg C<select>)
237
238=back
3a89a73c 239
0bdfc961
NC
240=head2 compressed man pages
241
242Be able to install them. This would probably need a configure test to see how
243the system does compressed man pages (same directory/different directory?
244same filename/different filename), as well as tweaking the F<installman> script
245to compress as necessary.
246
30222c0f
NC
247=head2 Add a code coverage target to the Makefile
248
249Make it easy for anyone to run Devel::Cover on the core's tests. The steps
250to do this manually are roughly
251
252=over 4
253
254=item *
255
256do a normal C<Configure>, but include Devel::Cover as a module to install
257(see F<INSTALL> for how to do this)
258
259=item *
260
261 make perl
262
263=item *
264
265 cd t; HARNESS_PERL_SWITCHES=-MDevel::Cover ./perl -I../lib harness
266
267=item *
268
269Process the resulting Devel::Cover database
270
271=back
272
273This just give you the coverage of the F<.pm>s. To also get the C level
274coverage you need to
275
276=over 4
277
278=item *
279
280Additionally tell C<Configure> to use the appropriate C compiler flags for
281C<gcov>
282
283=item *
284
285 make perl.gcov
286
287(instead of C<make perl>)
288
289=item *
290
291After running the tests run C<gcov> to generate all the F<.gcov> files.
292(Including down in the subdirectories of F<ext/>
293
294=item *
295
296(From the top level perl directory) run C<gcov2perl> on all the C<.gcov> files
297to get their stats into the cover_db directory.
298
299=item *
300
301Then process the Devel::Cover database
302
303=back
304
305It would be good to add a single switch to C<Configure> to specify that you
306wanted to perform perl level coverage, and another to specify C level
307coverage, and have C<Configure> and the F<Makefile> do all the right things
308automatically.
309
02f21748 310=head2 Make Config.pm cope with differences between built and installed perl
0bdfc961
NC
311
312Quite often vendors ship a perl binary compiled with their (pay-for)
313compilers. People install a free compiler, such as gcc. To work out how to
314build extensions, Perl interrogates C<%Config>, so in this situation
315C<%Config> describes compilers that aren't there, and extension building
316fails. This forces people into choosing between re-compiling perl themselves
317using the compiler they have, or only using modules that the vendor ships.
318
319It would be good to find a way teach C<Config.pm> about the installation setup,
320possibly involving probing at install time or later, so that the C<%Config> in
321a binary distribution better describes the installed machine, when the
322installed machine differs from the build machine in some significant way.
323
728f4ecd
NC
324=head2 linker specification files
325
326Some platforms mandate that you provide a list of a shared library's external
327symbols to the linker, so the core already has the infrastructure in place to
328do this for generating shared perl libraries. My understanding is that the
329GNU toolchain can accept an optional linker specification file, and restrict
330visibility just to symbols declared in that file. It would be good to extend
331F<makedef.pl> to support this format, and to provide a means within
332C<Configure> to enable it. This would allow Unix users to test that the
333export list is correct, and to build a perl that does not pollute the global
334namespace with private symbols.
335
a229ae3b
RGS
336=head2 Cross-compile support
337
338Currently C<Configure> understands C<-Dusecrosscompile> option. This option
339arranges for building C<miniperl> for TARGET machine, so this C<miniperl> is
340assumed then to be copied to TARGET machine and used as a replacement of full
341C<perl> executable.
342
d1307786 343This could be done little differently. Namely C<miniperl> should be built for
a229ae3b 344HOST and then full C<perl> with extensions should be compiled for TARGET.
d1307786 345This, however, might require extra trickery for %Config: we have one config
87a942b1
JH
346first for HOST and then another for TARGET. Tools like MakeMaker will be
347mightily confused. Having around two different types of executables and
348libraries (HOST and TARGET) makes life interesting for Makefiles and
349shell (and Perl) scripts. There is $Config{run}, normally empty, which
350can be used as an execution wrapper. Also note that in some
351cross-compilation/execution environments the HOST and the TARGET do
352not see the same filesystem(s), the $Config{run} may need to do some
353file/directory copying back and forth.
0bdfc961 354
8537f021
RGS
355=head2 roffitall
356
357Make F<pod/roffitall> be updated by F<pod/buildtoc>.
358
0bdfc961
NC
359=head1 Tasks that need a little C knowledge
360
361These tasks would need a little C knowledge, but don't need any specific
362background or experience with XS, or how the Perl interpreter works
363
3d826b29
NC
364=head2 Weed out needless PERL_UNUSED_ARG
365
366The C code uses the macro C<PERL_UNUSED_ARG> to stop compilers warning about
367unused arguments. Often the arguments can't be removed, as there is an
368external constraint that determines the prototype of the function, so this
369approach is valid. However, there are some cases where C<PERL_UNUSED_ARG>
370could be removed. Specifically
371
372=over 4
373
374=item *
375
376The prototypes of (nearly all) static functions can be changed
377
378=item *
379
380Unused arguments generated by short cut macros are wasteful - the short cut
381macro used can be changed.
382
383=back
384
fbf638cb
RGS
385=head2 Modernize the order of directories in @INC
386
387The way @INC is laid out by default, one cannot upgrade core (dual-life)
388modules without overwriting files. This causes problems for binary
3d14fd97
AD
389package builders. One possible proposal is laid out in this
390message:
391L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2002-04/msg02380.html>.
fbf638cb 392
bcbaa2d5
RGS
393=head2 -Duse32bit*
394
395Natively 64-bit systems need neither -Duse64bitint nor -Duse64bitall.
396On these systems, it might be the default compilation mode, and there
397is currently no guarantee that passing no use64bitall option to the
398Configure process will build a 32bit perl. Implementing -Duse32bit*
399options would be nice for perl 5.12.
400
0bdfc961 401=head2 Make it clear from -v if this is the exact official release
89007cb3 402
617eabfa
NC
403Currently perl from C<p4>/C<rsync> ships with a F<patchlevel.h> file that
404usually defines one local patch, of the form "MAINT12345" or "RC1". The output
405of perl -v doesn't report that a perl isn't an official release, and this
89007cb3 406information can get lost in bugs reports. Because of this, the minor version
fa11829f 407isn't bumped up until RC time, to minimise the possibility of versions of perl
89007cb3
NC
408escaping that believe themselves to be newer than they actually are.
409
410It would be useful to find an elegant way to have the "this is an interim
411maintenance release" or "this is a release candidate" in the terse -v output,
412and have it so that it's easy for the pumpking to remove this just as the
413release tarball is rolled up. This way the version pulled out of rsync would
414always say "I'm a development release" and it would be safe to bump the
415reported minor version as soon as a release ships, which would aid perl
416developers.
417
0bdfc961
NC
418This task is really about thinking of an elegant way to arrange the C source
419such that it's trivial for the Pumpking to flag "this is an official release"
420when making a tarball, yet leave the default source saying "I'm not the
421official release".
422
fee0a0f7 423=head2 Profile Perl - am I hot or not?
62403a3c 424
fee0a0f7
NC
425The Perl source code is stable enough that it makes sense to profile it,
426identify and optimise the hotspots. It would be good to measure the
427performance of the Perl interpreter using free tools such as cachegrind,
428gprof, and dtrace, and work to reduce the bottlenecks they reveal.
429
430As part of this, the idea of F<pp_hot.c> is that it contains the I<hot> ops,
431the ops that are most commonly used. The idea is that by grouping them, their
432object code will be adjacent in the executable, so they have a greater chance
433of already being in the CPU cache (or swapped in) due to being near another op
434already in use.
62403a3c
NC
435
436Except that it's not clear if these really are the most commonly used ops. So
fee0a0f7
NC
437as part of exercising your skills with coverage and profiling tools you might
438want to determine what ops I<really> are the most commonly used. And in turn
439suggest evictions and promotions to achieve a better F<pp_hot.c>.
62403a3c 440
91d0cbf6
NC
441One piece of Perl code that might make a good testbed is F<installman>.
442
98fed0ad
NC
443=head2 Allocate OPs from arenas
444
445Currently all new OP structures are individually malloc()ed and free()d.
446All C<malloc> implementations have space overheads, and are now as fast as
447custom allocates so it would both use less memory and less CPU to allocate
448the various OP structures from arenas. The SV arena code can probably be
449re-used for this.
450
539f2c54
JC
451Note that Configuring perl with C<-Accflags=-DPL_OP_SLAB_ALLOC> will use
452Perl_Slab_alloc() to pack optrees into a contiguous block, which is
453probably superior to the use of OP arenas, esp. from a cache locality
454standpoint. See L<Profile Perl - am I hot or not?>.
455
a229ae3b 456=head2 Improve win32/wince.c
0bdfc961 457
a229ae3b 458Currently, numerous functions look virtually, if not completely,
02f21748 459identical in both C<win32/wince.c> and C<win32/win32.c> files, which can't
6d71adcd
NC
460be good.
461
c5b31784
SH
462=head2 Use secure CRT functions when building with VC8 on Win32
463
464Visual C++ 2005 (VC++ 8.x) deprecated a number of CRT functions on the basis
465that they were "unsafe" and introduced differently named secure versions of
466them as replacements, e.g. instead of writing
467
468 FILE* f = fopen(__FILE__, "r");
469
470one should now write
471
472 FILE* f;
473 errno_t err = fopen_s(&f, __FILE__, "r");
474
475Currently, the warnings about these deprecations have been disabled by adding
476-D_CRT_SECURE_NO_DEPRECATE to the CFLAGS. It would be nice to remove that
477warning suppressant and actually make use of the new secure CRT functions.
478
479There is also a similar issue with POSIX CRT function names like fileno having
480been deprecated in favour of ISO C++ conformant names like _fileno. These
26a6faa8 481warnings are also currently suppressed by adding -D_CRT_NONSTDC_NO_DEPRECATE. It
c5b31784
SH
482might be nice to do as Microsoft suggest here too, although, unlike the secure
483functions issue, there is presumably little or no benefit in this case.
484
038ae9a4
SH
485=head2 Fix POSIX::access() and chdir() on Win32
486
487These functions currently take no account of DACLs and therefore do not behave
488correctly in situations where access is restricted by DACLs (as opposed to the
489read-only attribute).
490
491Furthermore, POSIX::access() behaves differently for directories having the
492read-only attribute set depending on what CRT library is being used. For
493example, the _access() function in the VC6 and VC7 CRTs (wrongly) claim that
494such directories are not writable, whereas in fact all directories are writable
495unless access is denied by DACLs. (In the case of directories, the read-only
496attribute actually only means that the directory cannot be deleted.) This CRT
497bug is fixed in the VC8 and VC9 CRTs (but, of course, the directory may still
498not actually be writable if access is indeed denied by DACLs).
499
500For the chdir() issue, see ActiveState bug #74552:
501http://bugs.activestate.com/show_bug.cgi?id=74552
502
503Therefore, DACLs should be checked both for consistency across CRTs and for
504the correct answer.
505
506(Note that perl's -w operator should not be modified to check DACLs. It has
507been written so that it reflects the state of the read-only attribute, even
508for directories (whatever CRT is being used), for symmetry with chmod().)
509
16815324
NC
510=head2 strcat(), strcpy(), strncat(), strncpy(), sprintf(), vsprintf()
511
512Maybe create a utility that checks after each libperl.a creation that
513none of the above (nor sprintf(), vsprintf(), or *SHUDDER* gets())
514ever creep back to libperl.a.
515
516 nm libperl.a | ./miniperl -alne '$o = $F[0] if /:$/; print "$o $F[1]" if $F[0] eq "U" && $F[1] =~ /^(?:strn?c(?:at|py)|v?sprintf|gets)$/'
517
518Note, of course, that this will only tell whether B<your> platform
519is using those naughty interfaces.
520
de96509d
JH
521=head2 -D_FORTIFY_SOURCE=2, -fstack-protector
522
523Recent glibcs support C<-D_FORTIFY_SOURCE=2> and recent gcc
524(4.1 onwards?) supports C<-fstack-protector>, both of which give
525protection against various kinds of buffer overflow problems.
526These should probably be used for compiling Perl whenever available,
527Configure and/or hints files should be adjusted to probe for the
528availability of these features and enable them as appropriate.
16815324 529
8964cfe0
NC
530=head2 Arenas for GPs? For MAGIC?
531
532C<struct gp> and C<struct magic> are both currently allocated by C<malloc>.
533It might be a speed or memory saving to change to using arenas. Or it might
534not. It would need some suitable benchmarking first. In particular, C<GP>s
535can probably be changed with minimal compatibility impact (probably nothing
536outside of the core, or even outside of F<gv.c> allocates them), but they
537probably aren't allocated/deallocated often enough for a speed saving. Whereas
538C<MAGIC> is allocated/deallocated more often, but in turn, is also something
539more externally visible, so changing the rules here may bite external code.
540
3880c8ec
NC
541=head2 Shared arenas
542
543Several SV body structs are now the same size, notably PVMG and PVGV, PVAV and
544PVHV, and PVCV and PVFM. It should be possible to allocate and return same
545sized bodies from the same actual arena, rather than maintaining one arena for
546each. This could save 4-6K per thread, of memory no longer tied up in the
547not-yet-allocated part of an arena.
548
8964cfe0 549
6d71adcd
NC
550=head1 Tasks that need a knowledge of XS
551
552These tasks would need C knowledge, and roughly the level of knowledge of
553the perl API that comes from writing modules that use XS to interface to
554C.
555
5d96f598
NC
556=head2 safely supporting POSIX SA_SIGINFO
557
558Some years ago Jarkko supplied patches to provide support for the POSIX
559SA_SIGINFO feature in Perl, passing the extra data to the Perl signal handler.
560
561Unfortunately, it only works with "unsafe" signals, because under safe
562signals, by the time Perl gets to run the signal handler, the extra
563information has been lost. Moreover, it's not easy to store it somewhere,
564as you can't call mutexs, or do anything else fancy, from inside a signal
565handler.
566
567So it strikes me that we could provide safe SA_SIGINFO support
568
569=over 4
570
571=item 1
572
573Provide global variables for two file descriptors
574
575=item 2
576
577When the first request is made via C<sigaction> for C<SA_SIGINFO>, create a
578pipe, store the reader in one, the writer in the other
579
580=item 3
581
582In the "safe" signal handler (C<Perl_csighandler()>/C<S_raise_signal()>), if
583the C<siginfo_t> pointer non-C<NULL>, and the writer file handle is open,
584
585=over 8
586
587=item 1
588
589serialise signal number, C<struct siginfo_t> (or at least the parts we care
590about) into a small auto char buff
591
592=item 2
593
594C<write()> that (non-blocking) to the writer fd
595
596=over 12
597
598=item 1
599
600if it writes 100%, flag the signal in a counter of "signals on the pipe" akin
601to the current per-signal-number counts
602
603=item 2
604
605if it writes 0%, assume the pipe is full. Flag the data as lost?
606
607=item 3
608
609if it writes partially, croak a panic, as your OS is broken.
610
611=back
612
613=back
614
615=item 4
616
617in the regular C<PERL_ASYNC_CHECK()> processing, if there are "signals on
618the pipe", read the data out, deserialise, build the Perl structures on
619the stack (code in C<Perl_sighandler()>, the "unsafe" handler), and call as
620usual.
621
622=back
623
624I think that this gets us decent C<SA_SIGINFO> support, without the current risk
625of running Perl code inside the signal handler context. (With all the dangers
626of things like C<malloc> corruption that that currently offers us)
627
628For more information see the thread starting with this message:
629http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-03/msg00305.html
630
6d71adcd
NC
631=head2 autovivification
632
633Make all autovivification consistent w.r.t LVALUE/RVALUE and strict/no strict;
634
635This task is incremental - even a little bit of work on it will help.
636
637=head2 Unicode in Filenames
638
639chdir, chmod, chown, chroot, exec, glob, link, lstat, mkdir, open,
640opendir, qx, readdir, readlink, rename, rmdir, stat, symlink, sysopen,
641system, truncate, unlink, utime, -X. All these could potentially accept
642Unicode filenames either as input or output (and in the case of system
643and qx Unicode in general, as input or output to/from the shell).
644Whether a filesystem - an operating system pair understands Unicode in
645filenames varies.
646
647Known combinations that have some level of understanding include
648Microsoft NTFS, Apple HFS+ (In Mac OS 9 and X) and Apple UFS (in Mac
649OS X), NFS v4 is rumored to be Unicode, and of course Plan 9. How to
650create Unicode filenames, what forms of Unicode are accepted and used
651(UCS-2, UTF-16, UTF-8), what (if any) is the normalization form used,
652and so on, varies. Finding the right level of interfacing to Perl
653requires some thought. Remember that an OS does not implicate a
654filesystem.
655
656(The Windows -C command flag "wide API support" has been at least
657temporarily retired in 5.8.1, and the -C has been repurposed, see
658L<perlrun>.)
659
87a942b1
JH
660Most probably the right way to do this would be this:
661L</"Virtualize operating system access">.
662
6d71adcd
NC
663=head2 Unicode in %ENV
664
665Currently the %ENV entries are always byte strings.
87a942b1 666See L</"Virtualize operating system access">.
6d71adcd 667
1f2e7916
JD
668=head2 Unicode and glob()
669
670Currently glob patterns and filenames returned from File::Glob::glob()
87a942b1 671are always byte strings. See L</"Virtualize operating system access">.
1f2e7916 672
dbb0c492
RGS
673=head2 Unicode and lc/uc operators
674
675Some built-in operators (C<lc>, C<uc>, etc.) behave differently, based on
676what the internal encoding of their argument is. That should not be the
677case. Maybe add a pragma to switch behaviour.
678
6d71adcd
NC
679=head2 use less 'memory'
680
681Investigate trade offs to switch out perl's choices on memory usage.
682Particularly perl should be able to give memory back.
683
684This task is incremental - even a little bit of work on it will help.
685
686=head2 Re-implement C<:unique> in a way that is actually thread-safe
687
688The old implementation made bad assumptions on several levels. A good 90%
689solution might be just to make C<:unique> work to share the string buffer
690of SvPVs. That way large constant strings can be shared between ithreads,
691such as the configuration information in F<Config>.
692
693=head2 Make tainting consistent
694
695Tainting would be easier to use if it didn't take documented shortcuts and
696allow taint to "leak" everywhere within an expression.
697
698=head2 readpipe(LIST)
699
700system() accepts a LIST syntax (and a PROGRAM LIST syntax) to avoid
701running a shell. readpipe() (the function behind qx//) could be similarly
702extended.
703
6d71adcd
NC
704=head2 Audit the code for destruction ordering assumptions
705
706Change 25773 notes
707
708 /* Need to check SvMAGICAL, as during global destruction it may be that
709 AvARYLEN(av) has been freed before av, and hence the SvANY() pointer
710 is now part of the linked list of SV heads, rather than pointing to
711 the original body. */
712 /* FIXME - audit the code for other bugs like this one. */
713
714adding the C<SvMAGICAL> check to
715
716 if (AvARYLEN(av) && SvMAGICAL(AvARYLEN(av))) {
717 MAGIC *mg = mg_find (AvARYLEN(av), PERL_MAGIC_arylen);
718
719Go through the core and look for similar assumptions that SVs have particular
720types, as all bets are off during global destruction.
721
749904bf
JH
722=head2 Extend PerlIO and PerlIO::Scalar
723
724PerlIO::Scalar doesn't know how to truncate(). Implementing this
725would require extending the PerlIO vtable.
726
727Similarly the PerlIO vtable doesn't know about formats (write()), or
728about stat(), or chmod()/chown(), utime(), or flock().
729
730(For PerlIO::Scalar it's hard to see what e.g. mode bits or ownership
731would mean.)
732
733PerlIO doesn't do directories or symlinks, either: mkdir(), rmdir(),
734opendir(), closedir(), seekdir(), rewinddir(), glob(); symlink(),
735readlink().
736
94da6c29
JH
737See also L</"Virtualize operating system access">.
738
3236f110
NC
739=head2 -C on the #! line
740
741It should be possible to make -C work correctly if found on the #! line,
742given that all perl command line options are strict ASCII, and -C changes
743only the interpretation of non-ASCII characters, and not for the script file
744handle. To make it work needs some investigation of the ordering of function
745calls during startup, and (by implication) a bit of tweaking of that order.
746
d6c1e11f
JH
747=head2 Organize error messages
748
749Perl's diagnostics (error messages, see L<perldiag>) could use
a8d0aeb9 750reorganizing and formalizing so that each error message has its
d6c1e11f
JH
751stable-for-all-eternity unique id, categorized by severity, type, and
752subsystem. (The error messages would be listed in a datafile outside
c4bd451b
CB
753of the Perl source code, and the source code would only refer to the
754messages by the id.) This clean-up and regularizing should apply
d6c1e11f
JH
755for all croak() messages.
756
757This would enable all sorts of things: easier translation/localization
758of the messages (though please do keep in mind the caveats of
759L<Locale::Maketext> about too straightforward approaches to
760translation), filtering by severity, and instead of grepping for a
761particular error message one could look for a stable error id. (Of
762course, changing the error messages by default would break all the
763existing software depending on some particular error message...)
764
765This kind of functionality is known as I<message catalogs>. Look for
766inspiration for example in the catgets() system, possibly even use it
767if available-- but B<only> if available, all platforms will B<not>
de96509d 768have catgets().
d6c1e11f
JH
769
770For the really pure at heart, consider extending this item to cover
771also the warning messages (see L<perllexwarn>, C<warnings.pl>).
3236f110 772
0bdfc961 773=head1 Tasks that need a knowledge of the interpreter
3298bd4d 774
0bdfc961
NC
775These tasks would need C knowledge, and knowledge of how the interpreter works,
776or a willingness to learn.
3298bd4d 777
718140ec
NC
778=head2 lexicals used only once
779
780This warns:
781
782 $ perl -we '$pie = 42'
783 Name "main::pie" used only once: possible typo at -e line 1.
784
785This does not:
786
787 $ perl -we 'my $pie = 42'
788
789Logically all lexicals used only once should warn, if the user asks for
d6f4ea2e
SP
790warnings. An unworked RT ticket (#5087) has been open for almost seven
791years for this discrepancy.
718140ec 792
a3d15f9a
RGS
793=head2 UTF-8 revamp
794
795The handling of Unicode is unclean in many places. For example, the regexp
796engine matches in Unicode semantics whenever the string or the pattern is
797flagged as UTF-8, but that should not be dependent on an internal storage
798detail of the string. Likewise, case folding behaviour is dependent on the
799UTF8 internal flag being on or off.
800
801=head2 Properly Unicode safe tokeniser and pads.
802
803The tokeniser isn't actually very UTF-8 clean. C<use utf8;> is a hack -
804variable names are stored in stashes as raw bytes, without the utf-8 flag
805set. The pad API only takes a C<char *> pointer, so that's all bytes too. The
806tokeniser ignores the UTF-8-ness of C<PL_rsfp>, or any SVs returned from
807source filters. All this could be fixed.
808
636e63cb
NC
809=head2 state variable initialization in list context
810
811Currently this is illegal:
812
813 state ($a, $b) = foo();
814
a2874905 815In Perl 6, C<state ($a) = foo();> and C<(state $a) = foo();> have different
a8d0aeb9 816semantics, which is tricky to implement in Perl 5 as currently they produce
a2874905 817the same opcode trees. The Perl 6 design is firm, so it would be good to
a8d0aeb9 818implement the necessary code in Perl 5. There are comments in
a2874905
NC
819C<Perl_newASSIGNOP()> that show the code paths taken by various assignment
820constructions involving state variables.
636e63cb 821
4fedb12c
RGS
822=head2 Implement $value ~~ 0 .. $range
823
824It would be nice to extend the syntax of the C<~~> operator to also
825understand numeric (and maybe alphanumeric) ranges.
a393eb28
RGS
826
827=head2 A does() built-in
828
829Like ref(), only useful. It would call the C<DOES> method on objects; it
830would also tell whether something can be dereferenced as an
831array/hash/etc., or used as a regexp, etc.
832L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-03/msg00481.html>
833
834=head2 Tied filehandles and write() don't mix
835
836There is no method on tied filehandles to allow them to be called back by
837formats.
4fedb12c 838
d10fc472 839=head2 Attach/detach debugger from running program
1626a787 840
cd793d32
NC
841The old perltodo notes "With C<gdb>, you can attach the debugger to a running
842program if you pass the process ID. It would be good to do this with the Perl
0bdfc961
NC
843debugger on a running Perl program, although I'm not sure how it would be
844done." ssh and screen do this with named pipes in /tmp. Maybe we can too.
1626a787 845
a8cb5b9e
RGS
846=head2 Optimize away empty destructors
847
848Defining an empty DESTROY method might be useful (notably in
849AUTOLOAD-enabled classes), but it's still a bit expensive to call. That
850could probably be optimized.
851
0bdfc961
NC
852=head2 LVALUE functions for lists
853
854The old perltodo notes that lvalue functions don't work for list or hash
855slices. This would be good to fix.
856
857=head2 LVALUE functions in the debugger
858
859The old perltodo notes that lvalue functions don't work in the debugger. This
860would be good to fix.
861
0bdfc961
NC
862=head2 regexp optimiser optional
863
864The regexp optimiser is not optional. It should configurable to be, to allow
865its performance to be measured, and its bugs to be easily demonstrated.
866
02f21748
RGS
867=head2 delete &function
868
869Allow to delete functions. One can already undef them, but they're still
870in the stash.
871
ef36c6a7
RGS
872=head2 C</w> regex modifier
873
874That flag would enable to match whole words, and also to interpolate
875arrays as alternations. With it, C</P/w> would be roughly equivalent to:
876
877 do { local $"='|'; /\b(?:P)\b/ }
878
879See L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-01/msg00400.html>
880for the discussion.
881
0bdfc961
NC
882=head2 optional optimizer
883
884Make the peephole optimizer optional. Currently it performs two tasks as
885it walks the optree - genuine peephole optimisations, and necessary fixups of
886ops. It would be good to find an efficient way to switch out the
887optimisations whilst keeping the fixups.
888
889=head2 You WANT *how* many
890
891Currently contexts are void, scalar and list. split has a special mechanism in
892place to pass in the number of return values wanted. It would be useful to
893have a general mechanism for this, backwards compatible and little speed hit.
894This would allow proposals such as short circuiting sort to be implemented
895as a module on CPAN.
896
897=head2 lexical aliases
898
899Allow lexical aliases (maybe via the syntax C<my \$alias = \$foo>.
900
901=head2 entersub XS vs Perl
902
903At the moment pp_entersub is huge, and has code to deal with entering both
904perl and XS subroutines. Subroutine implementations rarely change between
905perl and XS at run time, so investigate using 2 ops to enter subs (one for
906XS, one for perl) and swap between if a sub is redefined.
2810d901 907
de535794 908=head2 Self-ties
2810d901 909
de535794 910Self-ties are currently illegal because they caused too many segfaults. Maybe
a8d0aeb9 911the causes of these could be tracked down and self-ties on all types
de535794 912reinstated.
0bdfc961
NC
913
914=head2 Optimize away @_
915
916The old perltodo notes "Look at the "reification" code in C<av.c>".
917
87a942b1
JH
918=head2 Virtualize operating system access
919
920Implement a set of "vtables" that virtualizes operating system access
921(open(), mkdir(), unlink(), readdir(), getenv(), etc.) At the very
922least these interfaces should take SVs as "name" arguments instead of
923bare char pointers; probably the most flexible and extensible way
e1a3d5d1
JH
924would be for the Perl-facing interfaces to accept HVs. The system
925needs to be per-operating-system and per-file-system
926hookable/filterable, preferably both from XS and Perl level
87a942b1
JH
927(L<perlport/"Files and Filesystems"> is good reading at this point,
928in fact, all of L<perlport> is.)
929
e1a3d5d1
JH
930This has actually already been implemented (but only for Win32),
931take a look at F<iperlsys.h> and F<win32/perlhost.h>. While all Win32
932variants go through a set of "vtables" for operating system access,
933non-Win32 systems currently go straight for the POSIX/UNIX-style
934system/library call. Similar system as for Win32 should be
935implemented for all platforms. The existing Win32 implementation
936probably does not need to survive alongside this proposed new
937implementation, the approaches could be merged.
87a942b1
JH
938
939What would this give us? One often-asked-for feature this would
94da6c29
JH
940enable is using Unicode for filenames, and other "names" like %ENV,
941usernames, hostnames, and so forth.
942(See L<perlunicode/"When Unicode Does Not Happen">.)
943
944But this kind of virtualization would also allow for things like
945virtual filesystems, virtual networks, and "sandboxes" (though as long
946as dynamic loading of random object code is allowed, not very safe
947sandboxes since external code of course know not of Perl's vtables).
948An example of a smaller "sandbox" is that this feature can be used to
949implement per-thread working directories: Win32 already does this.
950
951See also L</"Extend PerlIO and PerlIO::Scalar">.
87a942b1 952
ac6197af
NC
953=head2 Investigate PADTMP hash pessimisation
954
955The peephole optimier converts constants used for hash key lookups to shared
057163d7 956hash key scalars. Under ithreads, something is undoing this work.
ac6197af
NC
957See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-09/msg00793.html
958
057163d7
NC
959=head2 Store the current pad in the OP slab allocator
960
961=for clarification
962I hope that I got that "current pad" part correct
963
964Currently we leak ops in various cases of parse failure. I suggested that we
965could solve this by always using the op slab allocator, and walking it to
966free ops. Dave comments that as some ops are already freed during optree
967creation one would have to mark which ops are freed, and not double free them
968when walking the slab. He notes that one problem with this is that for some ops
969you have to know which pad was current at the time of allocation, which does
970change. I suggested storing a pointer to the current pad in the memory allocated
971for the slab, and swapping to a new slab each time the pad changes. Dave thinks
972that this would work.
973
52960e22
JC
974=head2 repack the optree
975
976Repacking the optree after execution order is determined could allow
057163d7
NC
977removal of NULL ops, and optimal ordering of OPs with respect to cache-line
978filling. The slab allocator could be reused for this purpose. I think that
979the best way to do this is to make it an optional step just before the
980completed optree is attached to anything else, and to use the slab allocator
981unchanged, so that freeing ops is identical whether or not this step runs.
982Note that the slab allocator allocates ops downwards in memory, so one would
983have to actually "allocate" the ops in reverse-execution order to get them
984contiguous in memory in execution order.
985
986See http://www.nntp.perl.org/group/perl.perl5.porters/2007/12/msg131975.html
987
988Note that running this copy, and then freeing all the old location ops would
989cause their slabs to be freed, which would eliminate possible memory wastage if
990the previous suggestion is implemented, and we swap slabs more frequently.
52960e22 991
12e06b6f
NC
992=head2 eliminate incorrect line numbers in warnings
993
994This code
995
996 use warnings;
997 my $undef;
998
999 if ($undef == 3) {
1000 } elsif ($undef == 0) {
1001 }
1002
18a16cc5 1003used to produce this output:
12e06b6f
NC
1004
1005 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1006 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1007
18a16cc5
NC
1008where the line of the second warning was misreported - it should be line 5.
1009Rafael fixed this - the problem arose because there was no nextstate OP
1010between the execution of the C<if> and the C<elsif>, hence C<PL_curcop> still
1011reports that the currently executing line is line 4. The solution was to inject
1012a nextstate OPs for each C<elsif>, although it turned out that the nextstate
1013OP needed to be a nulled OP, rather than a live nextstate OP, else other line
1014numbers became misreported. (Jenga!)
12e06b6f
NC
1015
1016The problem is more general than C<elsif> (although the C<elsif> case is the
1017most common and the most confusing). Ideally this code
1018
1019 use warnings;
1020 my $undef;
1021
1022 my $a = $undef + 1;
1023 my $b
1024 = $undef
1025 + 1;
1026
1027would produce this output
1028
1029 Use of uninitialized value $undef in addition (+) at wrong.pl line 4.
1030 Use of uninitialized value $undef in addition (+) at wrong.pl line 7.
1031
1032(rather than lines 4 and 5), but this would seem to require every OP to carry
1033(at least) line number information.
1034
1035What might work is to have an optional line number in memory just before the
1036BASEOP structure, with a flag bit in the op to say whether it's present.
1037Initially during compile every OP would carry its line number. Then add a late
1038pass to the optimiser (potentially combined with L</repack the optree>) which
1039looks at the two ops on every edge of the graph of the execution path. If
1040the line number changes, flags the destination OP with this information.
1041Once all paths are traced, replace every op with the flag with a
1042nextstate-light op (that just updates C<PL_curcop>), which in turn then passes
1043control on to the true op. All ops would then be replaced by variants that
1044do not store the line number. (Which, logically, why it would work best in
1045conjunction with L</repack the optree>, as that is already copying/reallocating
1046all the OPs)
1047
18a16cc5
NC
1048(Although I should note that we're not certain that doing this for the general
1049case is worth it)
1050
52960e22
JC
1051=head2 optimize tail-calls
1052
1053Tail-calls present an opportunity for broadly applicable optimization;
1054anywhere that C<< return foo(...) >> is called, the outer return can
1055be replaced by a goto, and foo will return directly to the outer
1056caller, saving (conservatively) 25% of perl's call&return cost, which
1057is relatively higher than in C. The scheme language is known to do
1058this heavily. B::Concise provides good insight into where this
1059optimization is possible, ie anywhere entersub,leavesub op-sequence
1060occurs.
1061
1062 perl -MO=Concise,-exec,a,b,-main -e 'sub a{ 1 }; sub b {a()}; b(2)'
1063
1064Bottom line on this is probably a new pp_tailcall function which
1065combines the code in pp_entersub, pp_leavesub. This should probably
1066be done 1st in XS, and using B::Generate to patch the new OP into the
1067optrees.
1068
0bdfc961
NC
1069=head1 Big projects
1070
1071Tasks that will get your name mentioned in the description of the "Highlights
87a942b1 1072of 5.12"
0bdfc961
NC
1073
1074=head2 make ithreads more robust
1075
4e577f8b 1076Generally make ithreads more robust. See also L</iCOW>
0bdfc961
NC
1077
1078This task is incremental - even a little bit of work on it will help, and
1079will be greatly appreciated.
1080
6c047da7
YST
1081One bit would be to write the missing code in sv.c:Perl_dirp_dup.
1082
59c7f7d5
RGS
1083Fix Perl_sv_dup, et al so that threads can return objects.
1084
0bdfc961
NC
1085=head2 iCOW
1086
1087Sarathy and Arthur have a proposal for an improved Copy On Write which
1088specifically will be able to COW new ithreads. If this can be implemented
1089it would be a good thing.
1090
1091=head2 (?{...}) closures in regexps
1092
1093Fix (or rewrite) the implementation of the C</(?{...})/> closures.
1094
1095=head2 A re-entrant regexp engine
1096
1097This will allow the use of a regex from inside (?{ }), (??{ }) and
1098(?(?{ })|) constructs.
6bda09f9 1099
6bda09f9
YO
1100=head2 Add class set operations to regexp engine
1101
1102Apparently these are quite useful. Anyway, Jeffery Friedl wants them.
1103
1104demerphq has this on his todo list, but right at the bottom.