3 perltodo - Perl TO-DO List
7 This is a list of wishes for Perl. The most up to date version of this file
8 is at L<http://perl5.git.perl.org/perl.git/blob_plain/HEAD:/pod/perltodo.pod>
10 The tasks we think are smaller or easier are listed first. Anyone is welcome
11 to work on any of these, but it's a good idea to first contact
12 I<perl5-porters@perl.org> to avoid duplication of effort, and to learn from
13 any previous attempts. By all means contact a pumpking privately first if you
16 Whilst patches to make the list shorter are most welcome, ideas to add to
17 the list are also encouraged. Check the perl5-porters archives for past
18 ideas, and any discussion about them. One set of archives may be found at
19 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/>
21 What can we offer you in return? Fame, fortune, and everlasting glory? Maybe
22 not, but if your patch is incorporated, then we'll add your name to the
23 F<AUTHORS> file, which ships in the official distribution. How many other
24 programming languages offer you 1 line of immortality?
26 =head1 Tasks that only need Perl knowledge
28 =head2 Migrate t/ from custom TAP generation
30 Many tests below F<t/> still generate TAP by "hand", rather than using library
31 functions. As explained in L<perlhack/TESTING>, tests in F<t/> are
32 written in a particular way to test that more complex constructions actually
33 work before using them routinely. Hence they don't use C<Test::More>, but
34 instead there is an intentionally simpler library, F<t/test.pl>. However,
35 quite a few tests in F<t/> have not been refactored to use it. Refactoring
36 any of these tests, one at a time, is a useful thing TODO.
38 The subdirectories F<base>, F<cmd> and F<comp>, that contain the most
39 basic tests, should be excluded from this task.
41 =head2 Automate perldelta generation
43 The perldelta file accompanying each release summaries the major changes.
44 It's mostly manually generated currently, but some of that could be
45 automated with a bit of perl, specifically the generation of
49 =item Modules and Pragmata
51 =item New Documentation
57 See F<Porting/how_to_write_a_perldelta.pod> for details.
59 =head2 Remove duplication of test setup.
61 Schwern notes, that there's duplication of code - lots and lots of tests have
62 some variation on the big block of C<$Is_Foo> checks. We can safely put this
63 into a file, change it to build an C<%Is> hash and require it. Maybe just put
64 it into F<test.pl>. Throw in the handy tainting subroutines.
66 =head2 POD -E<gt> HTML conversion in the core still sucks
68 Which is crazy given just how simple POD purports to be, and how simple HTML
69 can be. It's not actually I<as> simple as it sounds, particularly with the
70 flexibility POD allows for C<=item>, but it would be good to improve the
71 visual appeal of the HTML generated, and to avoid it having any validation
72 errors. See also L</make HTML install work>, as the layout of installation tree
73 is needed to improve the cross-linking.
75 The addition of C<Pod::Simple> and its related modules may make this task
78 =head2 Make ExtUtils::ParseXS use strict;
80 F<lib/ExtUtils/ParseXS.pm> contains this line
82 # use strict; # One of these days...
84 Simply uncomment it, and fix all the resulting issues :-)
86 The more practical approach, to break the task down into manageable chunks, is
87 to work your way though the code from bottom to top, or if necessary adding
88 extra C<{ ... }> blocks, and turning on strict within them.
90 =head2 Make Schwern poorer
92 We should have tests for everything. When all the core's modules are tested,
93 Schwern has promised to donate to $500 to TPF. We may need volunteers to
94 hold him upside down and shake vigorously in order to actually extract the
97 =head2 Improve the coverage of the core tests
99 Use Devel::Cover to ascertain the core modules' test coverage, then add
100 tests that are currently missing.
104 A full test suite for the B module would be nice.
106 =head2 A decent benchmark
108 C<perlbench> seems impervious to any recent changes made to the perl core. It
109 would be useful to have a reasonable general benchmarking suite that roughly
110 represented what current perl programs do, and measurably reported whether
111 tweaks to the core improve, degrade or don't really affect performance, to
112 guide people attempting to optimise the guts of perl. Gisle would welcome
113 new tests for perlbench.
115 =head2 fix tainting bugs
117 Fix the bugs revealed by running the test suite with the C<-t> switch (via
118 C<make test.taintwarn>).
120 =head2 Dual life everything
122 As part of the "dists" plan, anything that doesn't belong in the smallest perl
123 distribution needs to be dual lifed. Anything else can be too. Figure out what
124 changes would be needed to package that module and its tests up for CPAN, and
125 do so. Test it with older perl releases, and fix the problems you find.
127 To make a minimal perl distribution, it's useful to look at
128 F<t/lib/commonsense.t>.
130 =head2 POSIX memory footprint
132 Ilya observed that use POSIX; eats memory like there's no tomorrow, and at
133 various times worked to cut it down. There is probably still fat to cut out -
134 for example POSIX passes Exporter some very memory hungry data structures.
136 =head2 embed.pl/makedef.pl
138 There is a script F<embed.pl> that generates several header files to prefix
139 all of Perl's symbols in a consistent way, to provide some semblance of
140 namespace support in C<C>. Functions are declared in F<embed.fnc>, variables
141 in F<interpvar.h>. Quite a few of the functions and variables
142 are conditionally declared there, using C<#ifdef>. However, F<embed.pl>
143 doesn't understand the C macros, so the rules about which symbols are present
144 when is duplicated in F<makedef.pl>. Writing things twice is bad, m'kay.
145 It would be good to teach C<embed.pl> to understand the conditional
146 compilation, and hence remove the duplication, and the mistakes it has caused.
148 =head2 use strict; and AutoLoad
150 Currently if you write
153 use AutoLoader 'AUTOLOAD';
158 print join (' ', No, strict, here), "!\n";
161 then C<use strict;> isn't in force within the autoloaded subroutines. It would
162 be more consistent (and less surprising) to arrange for all lexical pragmas
163 in force at the __END__ block to be in force within each autoloaded subroutine.
165 There's a similar problem with SelfLoader.
167 =head2 profile installman
169 The F<installman> script is slow. All it is doing text processing, which we're
170 told is something Perl is good at. So it would be nice to know what it is doing
171 that is taking so much CPU, and where possible address it.
173 =head2 enable lexical enabling/disabling of individual warnings
175 Currently, warnings can only be enabled or disabled by category. There
176 are times when it would be useful to quash a single warning, not a
179 =head2 document diagnostics
181 Many diagnostic messages are not currently documented. The list is at the end
184 =head1 Tasks that need a little sysadmin-type knowledge
186 Or if you prefer, tasks that you would learn from, and broaden your skills
189 =head2 make HTML install work
191 There is an C<installhtml> target in the Makefile. It's marked as
192 "experimental". It would be good to get this tested, make it work reliably, and
193 remove the "experimental" tag. This would include
199 Checking that cross linking between various parts of the documentation works.
200 In particular that links work between the modules (files with POD in F<lib/>)
201 and the core documentation (files in F<pod/>)
205 Work out how to split C<perlfunc> into chunks, preferably one per function
206 group, preferably with general case code that could be used elsewhere.
207 Challenges here are correctly identifying the groups of functions that go
208 together, and making the right named external cross-links point to the right
209 page. Things to be aware of are C<-X>, groups such as C<getpwnam> to
210 C<endservent>, two or more C<=items> giving the different parameter lists, such
213 =item substr EXPR,OFFSET,LENGTH,REPLACEMENT
214 =item substr EXPR,OFFSET,LENGTH
215 =item substr EXPR,OFFSET
217 and different parameter lists having different meanings. (eg C<select>)
221 =head2 compressed man pages
223 Be able to install them. This would probably need a configure test to see how
224 the system does compressed man pages (same directory/different directory?
225 same filename/different filename), as well as tweaking the F<installman> script
226 to compress as necessary.
228 =head2 Add a code coverage target to the Makefile
230 Make it easy for anyone to run Devel::Cover on the core's tests. The steps
231 to do this manually are roughly
237 do a normal C<Configure>, but include Devel::Cover as a module to install
238 (see L<INSTALL> for how to do this)
246 cd t; HARNESS_PERL_SWITCHES=-MDevel::Cover ./perl -I../lib harness
250 Process the resulting Devel::Cover database
254 This just give you the coverage of the F<.pm>s. To also get the C level
261 Additionally tell C<Configure> to use the appropriate C compiler flags for
268 (instead of C<make perl>)
272 After running the tests run C<gcov> to generate all the F<.gcov> files.
273 (Including down in the subdirectories of F<ext/>
277 (From the top level perl directory) run C<gcov2perl> on all the C<.gcov> files
278 to get their stats into the cover_db directory.
282 Then process the Devel::Cover database
286 It would be good to add a single switch to C<Configure> to specify that you
287 wanted to perform perl level coverage, and another to specify C level
288 coverage, and have C<Configure> and the F<Makefile> do all the right things
291 =head2 Make Config.pm cope with differences between built and installed perl
293 Quite often vendors ship a perl binary compiled with their (pay-for)
294 compilers. People install a free compiler, such as gcc. To work out how to
295 build extensions, Perl interrogates C<%Config>, so in this situation
296 C<%Config> describes compilers that aren't there, and extension building
297 fails. This forces people into choosing between re-compiling perl themselves
298 using the compiler they have, or only using modules that the vendor ships.
300 It would be good to find a way teach C<Config.pm> about the installation setup,
301 possibly involving probing at install time or later, so that the C<%Config> in
302 a binary distribution better describes the installed machine, when the
303 installed machine differs from the build machine in some significant way.
305 =head2 linker specification files
307 Some platforms mandate that you provide a list of a shared library's external
308 symbols to the linker, so the core already has the infrastructure in place to
309 do this for generating shared perl libraries. My understanding is that the
310 GNU toolchain can accept an optional linker specification file, and restrict
311 visibility just to symbols declared in that file. It would be good to extend
312 F<makedef.pl> to support this format, and to provide a means within
313 C<Configure> to enable it. This would allow Unix users to test that the
314 export list is correct, and to build a perl that does not pollute the global
315 namespace with private symbols, and will fail in the same way as msvc or mingw
316 builds or when using PERL_DL_NONLAZY=1.
318 =head2 Cross-compile support
320 Currently C<Configure> understands C<-Dusecrosscompile> option. This option
321 arranges for building C<miniperl> for TARGET machine, so this C<miniperl> is
322 assumed then to be copied to TARGET machine and used as a replacement of full
325 This could be done little differently. Namely C<miniperl> should be built for
326 HOST and then full C<perl> with extensions should be compiled for TARGET.
327 This, however, might require extra trickery for %Config: we have one config
328 first for HOST and then another for TARGET. Tools like MakeMaker will be
329 mightily confused. Having around two different types of executables and
330 libraries (HOST and TARGET) makes life interesting for Makefiles and
331 shell (and Perl) scripts. There is $Config{run}, normally empty, which
332 can be used as an execution wrapper. Also note that in some
333 cross-compilation/execution environments the HOST and the TARGET do
334 not see the same filesystem(s), the $Config{run} may need to do some
335 file/directory copying back and forth.
339 Make F<pod/roffitall> be updated by F<pod/buildtoc>.
341 =head2 Split "linker" from "compiler"
343 Right now, Configure probes for two commands, and sets two variables:
347 =item * C<cc> (in F<cc.U>)
349 This variable holds the name of a command to execute a C compiler which
350 can resolve multiple global references that happen to have the same
351 name. Usual values are F<cc> and F<gcc>.
352 Fervent ANSI compilers may be called F<c89>. AIX has F<xlc>.
354 =item * C<ld> (in F<dlsrc.U>)
356 This variable indicates the program to be used to link
357 libraries for dynamic loading. On some systems, it is F<ld>.
358 On ELF systems, it should be C<$cc>. Mostly, we'll try to respect
359 the hint file setting.
363 There is an implicit historical assumption from around Perl5.000alpha
364 something, that C<$cc> is also the correct command for linking object files
365 together to make an executable. This may be true on Unix, but it's not true
366 on other platforms, and there are a maze of work arounds in other places (such
367 as F<Makefile.SH>) to cope with this.
369 Ideally, we should create a new variable to hold the name of the executable
370 linker program, probe for it in F<Configure>, and centralise all the special
371 case logic there or in hints files.
373 A small bikeshed issue remains - what to call it, given that C<$ld> is already
374 taken (arguably for the wrong thing now, but on SunOS 4.1 it is the command
375 for creating dynamically-loadable modules) and C<$link> could be confused with
376 the Unix command line executable of the same name, which does something
377 completely different. Andy Dougherty makes the counter argument "In parrot, I
378 tried to call the command used to link object files and libraries into an
379 executable F<link>, since that's what my vaguely-remembered DOS and VMS
380 experience suggested. I don't think any real confusion has ensued, so it's
381 probably a reasonable name for perl5 to use."
383 "Alas, I've always worried that introducing it would make things worse,
384 since now the module building utilities would have to look for
385 C<$Config{link}> and institute a fall-back plan if it weren't found."
386 Although I can see that as confusing, given that C<$Config{d_link}> is true
387 when (hard) links are available.
389 =head2 Configure Windows using PowerShell
391 Currently, Windows uses hard-coded config files based to build the
392 config.h for compiling Perl. Makefiles are also hard-coded and need to be
393 hand edited prior to building Perl. While this makes it easy to create a perl.exe
394 that works across multiple Windows versions, being able to accurately
395 configure a perl.exe for a specific Windows versions and VS C++ would be
396 a nice enhancement. With PowerShell available on Windows XP and up, this
397 may now be possible. Step 1 might be to investigate whether this is possible
398 and use this to clean up our current makefile situation. Step 2 would be to
399 see if there would be a way to use our existing metaconfig units to configure a
400 Windows Perl or whether we go in a separate direction and make it so. Of
401 course, we all know what step 3 is.
403 =head1 Tasks that need a little C knowledge
405 These tasks would need a little C knowledge, but don't need any specific
406 background or experience with XS, or how the Perl interpreter works
408 =head2 Weed out needless PERL_UNUSED_ARG
410 The C code uses the macro C<PERL_UNUSED_ARG> to stop compilers warning about
411 unused arguments. Often the arguments can't be removed, as there is an
412 external constraint that determines the prototype of the function, so this
413 approach is valid. However, there are some cases where C<PERL_UNUSED_ARG>
414 could be removed. Specifically
420 The prototypes of (nearly all) static functions can be changed
424 Unused arguments generated by short cut macros are wasteful - the short cut
425 macro used can be changed.
429 =head2 Modernize the order of directories in @INC
431 The way @INC is laid out by default, one cannot upgrade core (dual-life)
432 modules without overwriting files. This causes problems for binary
433 package builders. One possible proposal is laid out in this
435 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2002-04/msg02380.html>
439 Natively 64-bit systems need neither -Duse64bitint nor -Duse64bitall.
440 On these systems, it might be the default compilation mode, and there
441 is currently no guarantee that passing no use64bitall option to the
442 Configure process will build a 32bit perl. Implementing -Duse32bit*
443 options would be nice for perl 5.14.
445 =head2 Profile Perl - am I hot or not?
447 The Perl source code is stable enough that it makes sense to profile it,
448 identify and optimise the hotspots. It would be good to measure the
449 performance of the Perl interpreter using free tools such as cachegrind,
450 gprof, and dtrace, and work to reduce the bottlenecks they reveal.
452 As part of this, the idea of F<pp_hot.c> is that it contains the I<hot> ops,
453 the ops that are most commonly used. The idea is that by grouping them, their
454 object code will be adjacent in the executable, so they have a greater chance
455 of already being in the CPU cache (or swapped in) due to being near another op
458 Except that it's not clear if these really are the most commonly used ops. So
459 as part of exercising your skills with coverage and profiling tools you might
460 want to determine what ops I<really> are the most commonly used. And in turn
461 suggest evictions and promotions to achieve a better F<pp_hot.c>.
463 One piece of Perl code that might make a good testbed is F<installman>.
465 =head2 Allocate OPs from arenas
467 Currently all new OP structures are individually malloc()ed and free()d.
468 All C<malloc> implementations have space overheads, and are now as fast as
469 custom allocates so it would both use less memory and less CPU to allocate
470 the various OP structures from arenas. The SV arena code can probably be
473 Note that Configuring perl with C<-Accflags=-DPL_OP_SLAB_ALLOC> will use
474 Perl_Slab_alloc() to pack optrees into a contiguous block, which is
475 probably superior to the use of OP arenas, esp. from a cache locality
476 standpoint. See L<Profile Perl - am I hot or not?>.
478 =head2 Improve win32/wince.c
480 Currently, numerous functions look virtually, if not completely,
481 identical in both F<win32/wince.c> and F<win32/win32.c> files, which can't
484 =head2 Use secure CRT functions when building with VC8 on Win32
486 Visual C++ 2005 (VC++ 8.x) deprecated a number of CRT functions on the basis
487 that they were "unsafe" and introduced differently named secure versions of
488 them as replacements, e.g. instead of writing
490 FILE* f = fopen(__FILE__, "r");
495 errno_t err = fopen_s(&f, __FILE__, "r");
497 Currently, the warnings about these deprecations have been disabled by adding
498 -D_CRT_SECURE_NO_DEPRECATE to the CFLAGS. It would be nice to remove that
499 warning suppressant and actually make use of the new secure CRT functions.
501 There is also a similar issue with POSIX CRT function names like fileno having
502 been deprecated in favour of ISO C++ conformant names like _fileno. These
503 warnings are also currently suppressed by adding -D_CRT_NONSTDC_NO_DEPRECATE. It
504 might be nice to do as Microsoft suggest here too, although, unlike the secure
505 functions issue, there is presumably little or no benefit in this case.
507 =head2 Fix POSIX::access() and chdir() on Win32
509 These functions currently take no account of DACLs and therefore do not behave
510 correctly in situations where access is restricted by DACLs (as opposed to the
511 read-only attribute).
513 Furthermore, POSIX::access() behaves differently for directories having the
514 read-only attribute set depending on what CRT library is being used. For
515 example, the _access() function in the VC6 and VC7 CRTs (wrongly) claim that
516 such directories are not writable, whereas in fact all directories are writable
517 unless access is denied by DACLs. (In the case of directories, the read-only
518 attribute actually only means that the directory cannot be deleted.) This CRT
519 bug is fixed in the VC8 and VC9 CRTs (but, of course, the directory may still
520 not actually be writable if access is indeed denied by DACLs).
522 For the chdir() issue, see ActiveState bug #74552:
523 L<http://bugs.activestate.com/show_bug.cgi?id=74552>
525 Therefore, DACLs should be checked both for consistency across CRTs and for
528 (Note that perl's -w operator should not be modified to check DACLs. It has
529 been written so that it reflects the state of the read-only attribute, even
530 for directories (whatever CRT is being used), for symmetry with chmod().)
532 =head2 strcat(), strcpy(), strncat(), strncpy(), sprintf(), vsprintf()
534 Maybe create a utility that checks after each libperl.a creation that
535 none of the above (nor sprintf(), vsprintf(), or *SHUDDER* gets())
536 ever creep back to libperl.a.
538 nm libperl.a | ./miniperl -alne '$o = $F[0] if /:$/; print "$o $F[1]" if $F[0] eq "U" && $F[1] =~ /^(?:strn?c(?:at|py)|v?sprintf|gets)$/'
540 Note, of course, that this will only tell whether B<your> platform
541 is using those naughty interfaces.
543 =head2 -D_FORTIFY_SOURCE=2, -fstack-protector
545 Recent glibcs support C<-D_FORTIFY_SOURCE=2> and recent gcc
546 (4.1 onwards?) supports C<-fstack-protector>, both of which give
547 protection against various kinds of buffer overflow problems.
548 These should probably be used for compiling Perl whenever available,
549 Configure and/or hints files should be adjusted to probe for the
550 availability of these features and enable them as appropriate.
552 =head2 Arenas for GPs? For MAGIC?
554 C<struct gp> and C<struct magic> are both currently allocated by C<malloc>.
555 It might be a speed or memory saving to change to using arenas. Or it might
556 not. It would need some suitable benchmarking first. In particular, C<GP>s
557 can probably be changed with minimal compatibility impact (probably nothing
558 outside of the core, or even outside of F<gv.c> allocates them), but they
559 probably aren't allocated/deallocated often enough for a speed saving. Whereas
560 C<MAGIC> is allocated/deallocated more often, but in turn, is also something
561 more externally visible, so changing the rules here may bite external code.
565 Several SV body structs are now the same size, notably PVMG and PVGV, PVAV and
566 PVHV, and PVCV and PVFM. It should be possible to allocate and return same
567 sized bodies from the same actual arena, rather than maintaining one arena for
568 each. This could save 4-6K per thread, of memory no longer tied up in the
569 not-yet-allocated part of an arena.
572 =head1 Tasks that need a knowledge of XS
574 These tasks would need C knowledge, and roughly the level of knowledge of
575 the perl API that comes from writing modules that use XS to interface to
578 =head2 Write an XS cookbook
580 Create pod/perlxscookbook.pod with short, task-focused 'recipes' in XS that
581 demonstrate common tasks and good practices. (Some of these might be
582 extracted from perlguts.) The target audience should be XS novices, who need
583 more examples than perlguts but something less overwhelming than perlapi.
584 Recipes should provide "one pretty good way to do it" instead of TIMTOWTDI.
586 Rather than focusing on interfacing Perl to C libraries, such a cookbook
587 should probably focus on how to optimize Perl routines by re-writing them
588 in XS. This will likely be more motivating to those who mostly work in
589 Perl but are looking to take the next step into XS.
591 Deconstructing and explaining some simpler XS modules could be one way to
592 bootstrap a cookbook. (List::Util? Class::XSAccessor? Tree::Ternary_XS?)
593 Another option could be deconstructing the implementation of some simpler
596 =head2 Allow XSUBs to inline themselves as OPs
598 For a simple XSUB, often the subroutine dispatch takes more time than the
599 XSUB itself. The tokeniser already has the ability to inline constant
600 subroutines - it would be good to provide a way to inline other subroutines.
602 Specifically, simplest approach looks to be to allow an XSUB to provide an
603 alternative implementation of itself as a custom OP. A new flag bit in
604 C<CvFLAGS()> would signal to the peephole optimiser to take an optree
607 b <@> leave[1 ref] vKP/REFC ->(end)
609 2 <;> nextstate(main 1 -e:1) v:{ ->3
610 a <2> sassign vKS/2 ->b
611 8 <1> entersub[t2] sKS/TARG,1 ->9
614 4 <$> const(IV 1) sM ->5
615 6 <1> rv2av[t1] lKM/1 ->7
617 - <1> ex-rv2cv sK ->-
618 7 <$> gv(*x) s/EARLYCV ->8
619 - <1> ex-rv2sv sKRM*/1 ->a
622 perform the symbol table lookup of C<rv2cv> and C<gv(*x)>, locate the
623 pointer to the custom OP that provides the direct implementation, and re-
624 write the optree something like:
626 b <@> leave[1 ref] vKP/REFC ->(end)
628 2 <;> nextstate(main 1 -e:1) v:{ ->3
629 a <2> sassign vKS/2 ->b
633 4 <$> const(IV 1) sM ->5
634 6 <1> rv2av[t1] lKM/1 ->7
636 - <1> ex-rv2cv sK ->-
637 - <$> ex-gv(*x) s/EARLYCV ->7
638 - <1> ex-rv2sv sKRM*/1 ->a
641 I<i.e.> the C<gv(*)> OP has been nulled and spliced out of the execution
642 path, and the C<entersub> OP has been replaced by the custom op.
644 This approach should provide a measurable speed up to simple XSUBs inside
645 tight loops. Initially one would have to write the OP alternative
646 implementation by hand, but it's likely that this should be reasonably
647 straightforward for the type of XSUB that would benefit the most. Longer
648 term, once the run-time implementation is proven, it should be possible to
649 progressively update ExtUtils::ParseXS to generate OP implementations for
652 =head2 Remove the use of SVs as temporaries in dump.c
654 F<dump.c> contains debugging routines to dump out the contains of perl data
655 structures, such as C<SV>s, C<AV>s and C<HV>s. Currently, the dumping code
656 B<uses> C<SV>s for its temporary buffers, which was a logical initial
657 implementation choice, as they provide ready made memory handling.
659 However, they also lead to a lot of confusion when it happens that what you're
660 trying to debug is seen by the code in F<dump.c>, correctly or incorrectly, as
661 a temporary scalar it can use for a temporary buffer. It's also not possible
662 to dump scalars before the interpreter is properly set up, such as during
663 ithreads cloning. It would be good to progressively replace the use of scalars
664 as string accumulation buffers with something much simpler, directly allocated
665 by C<malloc>. The F<dump.c> code is (or should be) only producing 7 bit
666 US-ASCII, so output character sets are not an issue.
668 Producing and proving an internal simple buffer allocation would make it easier
669 to re-write the internals of the PerlIO subsystem to avoid using C<SV>s for
670 B<its> buffers, use of which can cause problems similar to those of F<dump.c>,
673 =head2 safely supporting POSIX SA_SIGINFO
675 Some years ago Jarkko supplied patches to provide support for the POSIX
676 SA_SIGINFO feature in Perl, passing the extra data to the Perl signal handler.
678 Unfortunately, it only works with "unsafe" signals, because under safe
679 signals, by the time Perl gets to run the signal handler, the extra
680 information has been lost. Moreover, it's not easy to store it somewhere,
681 as you can't call mutexs, or do anything else fancy, from inside a signal
684 So it strikes me that we could provide safe SA_SIGINFO support
690 Provide global variables for two file descriptors
694 When the first request is made via C<sigaction> for C<SA_SIGINFO>, create a
695 pipe, store the reader in one, the writer in the other
699 In the "safe" signal handler (C<Perl_csighandler()>/C<S_raise_signal()>), if
700 the C<siginfo_t> pointer non-C<NULL>, and the writer file handle is open,
706 serialise signal number, C<struct siginfo_t> (or at least the parts we care
707 about) into a small auto char buff
711 C<write()> that (non-blocking) to the writer fd
717 if it writes 100%, flag the signal in a counter of "signals on the pipe" akin
718 to the current per-signal-number counts
722 if it writes 0%, assume the pipe is full. Flag the data as lost?
726 if it writes partially, croak a panic, as your OS is broken.
734 in the regular C<PERL_ASYNC_CHECK()> processing, if there are "signals on
735 the pipe", read the data out, deserialise, build the Perl structures on
736 the stack (code in C<Perl_sighandler()>, the "unsafe" handler), and call as
741 I think that this gets us decent C<SA_SIGINFO> support, without the current risk
742 of running Perl code inside the signal handler context. (With all the dangers
743 of things like C<malloc> corruption that that currently offers us)
745 For more information see the thread starting with this message:
746 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-03/msg00305.html>
748 =head2 autovivification
750 Make all autovivification consistent w.r.t LVALUE/RVALUE and strict/no strict;
752 This task is incremental - even a little bit of work on it will help.
754 =head2 Unicode in Filenames
756 chdir, chmod, chown, chroot, exec, glob, link, lstat, mkdir, open,
757 opendir, qx, readdir, readlink, rename, rmdir, stat, symlink, sysopen,
758 system, truncate, unlink, utime, -X. All these could potentially accept
759 Unicode filenames either as input or output (and in the case of system
760 and qx Unicode in general, as input or output to/from the shell).
761 Whether a filesystem - an operating system pair understands Unicode in
764 Known combinations that have some level of understanding include
765 Microsoft NTFS, Apple HFS+ (In Mac OS 9 and X) and Apple UFS (in Mac
766 OS X), NFS v4 is rumored to be Unicode, and of course Plan 9. How to
767 create Unicode filenames, what forms of Unicode are accepted and used
768 (UCS-2, UTF-16, UTF-8), what (if any) is the normalization form used,
769 and so on, varies. Finding the right level of interfacing to Perl
770 requires some thought. Remember that an OS does not implicate a
773 (The Windows -C command flag "wide API support" has been at least
774 temporarily retired in 5.8.1, and the -C has been repurposed, see
777 Most probably the right way to do this would be this:
778 L</"Virtualize operating system access">.
780 =head2 Unicode in %ENV
782 Currently the %ENV entries are always byte strings.
783 See L</"Virtualize operating system access">.
785 =head2 Unicode and glob()
787 Currently glob patterns and filenames returned from File::Glob::glob()
788 are always byte strings. See L</"Virtualize operating system access">.
790 =head2 use less 'memory'
792 Investigate trade offs to switch out perl's choices on memory usage.
793 Particularly perl should be able to give memory back.
795 This task is incremental - even a little bit of work on it will help.
797 =head2 Re-implement C<:unique> in a way that is actually thread-safe
799 The old implementation made bad assumptions on several levels. A good 90%
800 solution might be just to make C<:unique> work to share the string buffer
801 of SvPVs. That way large constant strings can be shared between ithreads,
802 such as the configuration information in F<Config>.
804 =head2 Make tainting consistent
806 Tainting would be easier to use if it didn't take documented shortcuts and
807 allow taint to "leak" everywhere within an expression.
809 =head2 readpipe(LIST)
811 system() accepts a LIST syntax (and a PROGRAM LIST syntax) to avoid
812 running a shell. readpipe() (the function behind qx//) could be similarly
815 =head2 Audit the code for destruction ordering assumptions
819 /* Need to check SvMAGICAL, as during global destruction it may be that
820 AvARYLEN(av) has been freed before av, and hence the SvANY() pointer
821 is now part of the linked list of SV heads, rather than pointing to
822 the original body. */
823 /* FIXME - audit the code for other bugs like this one. */
825 adding the C<SvMAGICAL> check to
827 if (AvARYLEN(av) && SvMAGICAL(AvARYLEN(av))) {
828 MAGIC *mg = mg_find (AvARYLEN(av), PERL_MAGIC_arylen);
830 Go through the core and look for similar assumptions that SVs have particular
831 types, as all bets are off during global destruction.
833 =head2 Extend PerlIO and PerlIO::Scalar
835 PerlIO::Scalar doesn't know how to truncate(). Implementing this
836 would require extending the PerlIO vtable.
838 Similarly the PerlIO vtable doesn't know about formats (write()), or
839 about stat(), or chmod()/chown(), utime(), or flock().
841 (For PerlIO::Scalar it's hard to see what e.g. mode bits or ownership
844 PerlIO doesn't do directories or symlinks, either: mkdir(), rmdir(),
845 opendir(), closedir(), seekdir(), rewinddir(), glob(); symlink(),
848 See also L</"Virtualize operating system access">.
850 =head2 -C on the #! line
852 It should be possible to make -C work correctly if found on the #! line,
853 given that all perl command line options are strict ASCII, and -C changes
854 only the interpretation of non-ASCII characters, and not for the script file
855 handle. To make it work needs some investigation of the ordering of function
856 calls during startup, and (by implication) a bit of tweaking of that order.
858 =head2 Organize error messages
860 Perl's diagnostics (error messages, see L<perldiag>) could use
861 reorganizing and formalizing so that each error message has its
862 stable-for-all-eternity unique id, categorized by severity, type, and
863 subsystem. (The error messages would be listed in a datafile outside
864 of the Perl source code, and the source code would only refer to the
865 messages by the id.) This clean-up and regularizing should apply
866 for all croak() messages.
868 This would enable all sorts of things: easier translation/localization
869 of the messages (though please do keep in mind the caveats of
870 L<Locale::Maketext> about too straightforward approaches to
871 translation), filtering by severity, and instead of grepping for a
872 particular error message one could look for a stable error id. (Of
873 course, changing the error messages by default would break all the
874 existing software depending on some particular error message...)
876 This kind of functionality is known as I<message catalogs>. Look for
877 inspiration for example in the catgets() system, possibly even use it
878 if available-- but B<only> if available, all platforms will B<not>
881 For the really pure at heart, consider extending this item to cover
882 also the warning messages (see L<perllexwarn>, C<warnings.pl>).
884 =head1 Tasks that need a knowledge of the interpreter
886 These tasks would need C knowledge, and knowledge of how the interpreter works,
887 or a willingness to learn.
889 =head2 forbid labels with keyword names
891 Currently C<goto keyword> "computes" the label value:
893 $ perl -e 'goto print'
894 Can't find label 1 at -e line 1.
896 It is controversial if the right way to avoid the confusion is to forbid
897 labels with keyword names, or if it would be better to always treat
898 bareword expressions after a "goto" as a label and never as a keyword.
900 =head2 truncate() prototype
902 The prototype of truncate() is currently C<$$>. It should probably
903 be C<*$> instead. (This is changed in F<opcode.pl>)
905 =head2 decapsulation of smart match argument
907 Currently C<$foo ~~ $object> will die with the message "Smart matching a
908 non-overloaded object breaks encapsulation". It would be nice to allow
909 to bypass this by using explicitly the syntax C<$foo ~~ %$object> or
912 =head2 error reporting of [$a ; $b]
914 Using C<;> inside brackets is a syntax error, and we don't propose to change
915 that by giving it any meaning. However, it's not reported very helpfully:
917 $ perl -e '$a = [$b; $c];'
918 syntax error at -e line 1, near "$b;"
919 syntax error at -e line 1, near "$c]"
920 Execution of -e aborted due to compilation errors.
922 It should be possible to hook into the tokeniser or the lexer, so that when a
923 C<;> is parsed where it is not legal as a statement terminator (ie inside
924 C<{}> used as a hashref, C<[]> or C<()>) it issues an error something like
925 I<';' isn't legal inside an expression - if you need multiple statements use a
926 do {...} block>. See the thread starting at
927 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-09/msg00573.html>
929 =head2 lexicals used only once
933 $ perl -we '$pie = 42'
934 Name "main::pie" used only once: possible typo at -e line 1.
938 $ perl -we 'my $pie = 42'
940 Logically all lexicals used only once should warn, if the user asks for
941 warnings. An unworked RT ticket (#5087) has been open for almost seven
942 years for this discrepancy.
946 The handling of Unicode is unclean in many places. In the regex engine
947 there are especially many problems. The swash data structure could be
948 replaced my something better. Inversion lists and maps are likely
949 candidates. The whole Unicode database could be placed in-core for a
950 huge speed-up. Only minimal work was done on the optimizer when utf8
951 was added, with the result that the synthetic start class often will
952 fail to narrow down the possible choices when given non-Latin1 input.
954 =head2 Properly Unicode safe tokeniser and pads.
956 The tokeniser isn't actually very UTF-8 clean. C<use utf8;> is a hack -
957 variable names are stored in stashes as raw bytes, without the utf-8 flag
958 set. The pad API only takes a C<char *> pointer, so that's all bytes too. The
959 tokeniser ignores the UTF-8-ness of C<PL_rsfp>, or any SVs returned from
960 source filters. All this could be fixed.
962 =head2 state variable initialization in list context
964 Currently this is illegal:
966 state ($a, $b) = foo();
968 In Perl 6, C<state ($a) = foo();> and C<(state $a) = foo();> have different
969 semantics, which is tricky to implement in Perl 5 as currently they produce
970 the same opcode trees. The Perl 6 design is firm, so it would be good to
971 implement the necessary code in Perl 5. There are comments in
972 C<Perl_newASSIGNOP()> that show the code paths taken by various assignment
973 constructions involving state variables.
975 =head2 Implement $value ~~ 0 .. $range
977 It would be nice to extend the syntax of the C<~~> operator to also
978 understand numeric (and maybe alphanumeric) ranges.
980 =head2 A does() built-in
982 Like ref(), only useful. It would call the C<DOES> method on objects; it
983 would also tell whether something can be dereferenced as an
984 array/hash/etc., or used as a regexp, etc.
985 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-03/msg00481.html>
987 =head2 Tied filehandles and write() don't mix
989 There is no method on tied filehandles to allow them to be called back by
992 =head2 Propagate compilation hints to the debugger
994 Currently a debugger started with -dE on the command-line doesn't see the
995 features enabled by -E. More generally hints (C<$^H> and C<%^H>) aren't
996 propagated to the debugger. Probably it would be a good thing to propagate
997 hints from the innermost non-C<DB::> scope: this would make code eval'ed
998 in the debugger see the features (and strictures, etc.) currently in
1001 =head2 Attach/detach debugger from running program
1003 The old perltodo notes "With C<gdb>, you can attach the debugger to a running
1004 program if you pass the process ID. It would be good to do this with the Perl
1005 debugger on a running Perl program, although I'm not sure how it would be
1006 done." ssh and screen do this with named pipes in /tmp. Maybe we can too.
1008 =head2 LVALUE functions for lists
1010 The old perltodo notes that lvalue functions don't work for list or hash
1011 slices. This would be good to fix.
1013 =head2 regexp optimiser optional
1015 The regexp optimiser is not optional. It should configurable to be, to allow
1016 its performance to be measured, and its bugs to be easily demonstrated.
1018 =head2 C</w> regex modifier
1020 That flag would enable to match whole words, and also to interpolate
1021 arrays as alternations. With it, C</P/w> would be roughly equivalent to:
1023 do { local $"='|'; /\b(?:P)\b/ }
1026 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-01/msg00400.html>
1029 =head2 optional optimizer
1031 Make the peephole optimizer optional. Currently it performs two tasks as
1032 it walks the optree - genuine peephole optimisations, and necessary fixups of
1033 ops. It would be good to find an efficient way to switch out the
1034 optimisations whilst keeping the fixups.
1036 =head2 You WANT *how* many
1038 Currently contexts are void, scalar and list. split has a special mechanism in
1039 place to pass in the number of return values wanted. It would be useful to
1040 have a general mechanism for this, backwards compatible and little speed hit.
1041 This would allow proposals such as short circuiting sort to be implemented
1042 as a module on CPAN.
1044 =head2 lexical aliases
1046 Allow lexical aliases (maybe via the syntax C<my \$alias = \$foo>.
1048 =head2 entersub XS vs Perl
1050 At the moment pp_entersub is huge, and has code to deal with entering both
1051 perl and XS subroutines. Subroutine implementations rarely change between
1052 perl and XS at run time, so investigate using 2 ops to enter subs (one for
1053 XS, one for perl) and swap between if a sub is redefined.
1057 Self-ties are currently illegal because they caused too many segfaults. Maybe
1058 the causes of these could be tracked down and self-ties on all types
1061 =head2 Optimize away @_
1063 The old perltodo notes "Look at the "reification" code in C<av.c>".
1065 =head2 Virtualize operating system access
1067 Implement a set of "vtables" that virtualizes operating system access
1068 (open(), mkdir(), unlink(), readdir(), getenv(), etc.) At the very
1069 least these interfaces should take SVs as "name" arguments instead of
1070 bare char pointers; probably the most flexible and extensible way
1071 would be for the Perl-facing interfaces to accept HVs. The system
1072 needs to be per-operating-system and per-file-system
1073 hookable/filterable, preferably both from XS and Perl level
1074 (L<perlport/"Files and Filesystems"> is good reading at this point,
1075 in fact, all of L<perlport> is.)
1077 This has actually already been implemented (but only for Win32),
1078 take a look at F<iperlsys.h> and F<win32/perlhost.h>. While all Win32
1079 variants go through a set of "vtables" for operating system access,
1080 non-Win32 systems currently go straight for the POSIX/Unix-style
1081 system/library call. Similar system as for Win32 should be
1082 implemented for all platforms. The existing Win32 implementation
1083 probably does not need to survive alongside this proposed new
1084 implementation, the approaches could be merged.
1086 What would this give us? One often-asked-for feature this would
1087 enable is using Unicode for filenames, and other "names" like %ENV,
1088 usernames, hostnames, and so forth.
1089 (See L<perlunicode/"When Unicode Does Not Happen">.)
1091 But this kind of virtualization would also allow for things like
1092 virtual filesystems, virtual networks, and "sandboxes" (though as long
1093 as dynamic loading of random object code is allowed, not very safe
1094 sandboxes since external code of course know not of Perl's vtables).
1095 An example of a smaller "sandbox" is that this feature can be used to
1096 implement per-thread working directories: Win32 already does this.
1098 See also L</"Extend PerlIO and PerlIO::Scalar">.
1100 =head2 Investigate PADTMP hash pessimisation
1102 The peephole optimiser converts constants used for hash key lookups to shared
1103 hash key scalars. Under ithreads, something is undoing this work.
1105 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-09/msg00793.html>
1107 =head2 Store the current pad in the OP slab allocator
1110 I hope that I got that "current pad" part correct
1112 Currently we leak ops in various cases of parse failure. I suggested that we
1113 could solve this by always using the op slab allocator, and walking it to
1114 free ops. Dave comments that as some ops are already freed during optree
1115 creation one would have to mark which ops are freed, and not double free them
1116 when walking the slab. He notes that one problem with this is that for some ops
1117 you have to know which pad was current at the time of allocation, which does
1118 change. I suggested storing a pointer to the current pad in the memory allocated
1119 for the slab, and swapping to a new slab each time the pad changes. Dave thinks
1120 that this would work.
1122 =head2 repack the optree
1124 Repacking the optree after execution order is determined could allow
1125 removal of NULL ops, and optimal ordering of OPs with respect to cache-line
1126 filling. The slab allocator could be reused for this purpose. I think that
1127 the best way to do this is to make it an optional step just before the
1128 completed optree is attached to anything else, and to use the slab allocator
1129 unchanged, so that freeing ops is identical whether or not this step runs.
1130 Note that the slab allocator allocates ops downwards in memory, so one would
1131 have to actually "allocate" the ops in reverse-execution order to get them
1132 contiguous in memory in execution order.
1135 L<http://www.nntp.perl.org/group/perl.perl5.porters/2007/12/msg131975.html>
1137 Note that running this copy, and then freeing all the old location ops would
1138 cause their slabs to be freed, which would eliminate possible memory wastage if
1139 the previous suggestion is implemented, and we swap slabs more frequently.
1141 =head2 eliminate incorrect line numbers in warnings
1149 } elsif ($undef == 0) {
1152 used to produce this output:
1154 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1155 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1157 where the line of the second warning was misreported - it should be line 5.
1158 Rafael fixed this - the problem arose because there was no nextstate OP
1159 between the execution of the C<if> and the C<elsif>, hence C<PL_curcop> still
1160 reports that the currently executing line is line 4. The solution was to inject
1161 a nextstate OPs for each C<elsif>, although it turned out that the nextstate
1162 OP needed to be a nulled OP, rather than a live nextstate OP, else other line
1163 numbers became misreported. (Jenga!)
1165 The problem is more general than C<elsif> (although the C<elsif> case is the
1166 most common and the most confusing). Ideally this code
1176 would produce this output
1178 Use of uninitialized value $undef in addition (+) at wrong.pl line 4.
1179 Use of uninitialized value $undef in addition (+) at wrong.pl line 7.
1181 (rather than lines 4 and 5), but this would seem to require every OP to carry
1182 (at least) line number information.
1184 What might work is to have an optional line number in memory just before the
1185 BASEOP structure, with a flag bit in the op to say whether it's present.
1186 Initially during compile every OP would carry its line number. Then add a late
1187 pass to the optimiser (potentially combined with L</repack the optree>) which
1188 looks at the two ops on every edge of the graph of the execution path. If
1189 the line number changes, flags the destination OP with this information.
1190 Once all paths are traced, replace every op with the flag with a
1191 nextstate-light op (that just updates C<PL_curcop>), which in turn then passes
1192 control on to the true op. All ops would then be replaced by variants that
1193 do not store the line number. (Which, logically, why it would work best in
1194 conjunction with L</repack the optree>, as that is already copying/reallocating
1197 (Although I should note that we're not certain that doing this for the general
1200 =head2 optimize tail-calls
1202 Tail-calls present an opportunity for broadly applicable optimization;
1203 anywhere that C<< return foo(...) >> is called, the outer return can
1204 be replaced by a goto, and foo will return directly to the outer
1205 caller, saving (conservatively) 25% of perl's call&return cost, which
1206 is relatively higher than in C. The scheme language is known to do
1207 this heavily. B::Concise provides good insight into where this
1208 optimization is possible, ie anywhere entersub,leavesub op-sequence
1211 perl -MO=Concise,-exec,a,b,-main -e 'sub a{ 1 }; sub b {a()}; b(2)'
1213 Bottom line on this is probably a new pp_tailcall function which
1214 combines the code in pp_entersub, pp_leavesub. This should probably
1215 be done 1st in XS, and using B::Generate to patch the new OP into the
1218 =head2 Add C<00dddd>
1220 It has been proposed that octal constants be specifiable through the syntax
1221 C<0oddddd>, parallel to the existing construct to specify hex constants
1226 Tasks that will get your name mentioned in the description of the "Highlights
1229 =head2 make ithreads more robust
1231 Generally make ithreads more robust. See also L</iCOW>
1233 This task is incremental - even a little bit of work on it will help, and
1234 will be greatly appreciated.
1236 One bit would be to determine how to clone directory handles on systems
1237 without a C<fchdir> function (in sv.c:Perl_dirp_dup).
1239 Fix Perl_sv_dup, et al so that threads can return objects.
1243 Sarathy and Arthur have a proposal for an improved Copy On Write which
1244 specifically will be able to COW new ithreads. If this can be implemented
1245 it would be a good thing.
1247 =head2 (?{...}) closures in regexps
1249 Fix (or rewrite) the implementation of the C</(?{...})/> closures.
1251 =head2 Add class set operations to regexp engine
1253 Apparently these are quite useful. Anyway, Jeffery Friedl wants them.
1255 demerphq has this on his todo list, but right at the bottom.
1258 =head1 Tasks for microperl
1261 [ Each and every one of these may be obsolete, but they were listed
1262 in the old Todo.micro file]
1265 =head2 make creating uconfig.sh automatic
1267 =head2 make creating Makefile.micro automatic
1269 =head2 do away with fork/exec/wait?
1271 (system, popen should be enough?)
1273 =head2 some of the uconfig.sh really needs to be probed (using cc) in buildtime:
1275 (uConfigure? :-) native datatype widths and endianness come to mind