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 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:
20 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/
22 What can we offer you in return? Fame, fortune, and everlasting glory? Maybe
23 not, but if your patch is incorporated, then we'll add your name to the
24 F<AUTHORS> file, which ships in the official distribution. How many other
25 programming languages offer you 1 line of immortality?
27 =head1 Tasks that only need Perl knowledge
29 =head2 Migrate t/ from custom TAP generation
31 Many tests below F<t/> still generate TAP by "hand", rather than using library
32 functions. As explained in L<perlhack/Writing a test>, tests in F<t/> are
33 written in a particular way to test that more complex constructions actually
34 work before using them routinely. Hence they don't use C<Test::More>, but
35 instead there is an intentionally simpler library, F<t/test.pl>. However,
36 quite a few tests in F<t/> have not been refactored to use it. Refactoring
37 any of these tests, one at a time, is a useful thing TODO.
39 The subdirectories F<base>, F<cmd> and F<comp>, that contain the most
40 basic tests, should be excluded from this task.
42 =head2 Test that regen.pl was run
44 There are various generated files shipped with the perl distribution, for
45 things like header files generate from data. The generation scripts are
46 written in perl, and all can be run by F<regen.pl>. However, because they're
47 written in perl, we can't run them before we've built perl. We can't run them
48 as part of the F<Makefile>, because changing files underneath F<make> confuses
49 it completely, and we don't want to run them automatically anyway, as they
50 change files shipped by the distribution, something we seek not do to.
52 If someone changes the data, but forgets to re-run F<regen.pl> then the
53 generated files are out of sync. It would be good to have a test in
54 F<t/porting> that checks that the generated files are in sync, and fails
55 otherwise, to alert someone before they make a poor commit. I suspect that this
56 would require adapting the scripts run from F<regen.pl> to have dry-run
57 options, and invoking them with these, or by refactoring them into a library
58 that does the generation, which can be called by the scripts, and by the test.
60 =head2 Automate perldelta generation
62 The perldelta file accompanying each release summaries the major changes.
63 It's mostly manually generated currently, but some of that could be
64 automated with a bit of perl, specifically the generation of
68 =item Modules and Pragmata
70 =item New Documentation
76 See F<Porting/how_to_write_a_perldelta.pod> for details.
78 =head2 Remove duplication of test setup.
80 Schwern notes, that there's duplication of code - lots and lots of tests have
81 some variation on the big block of C<$Is_Foo> checks. We can safely put this
82 into a file, change it to build an C<%Is> hash and require it. Maybe just put
83 it into F<test.pl>. Throw in the handy tainting subroutines.
85 =head2 POD -E<gt> HTML conversion in the core still sucks
87 Which is crazy given just how simple POD purports to be, and how simple HTML
88 can be. It's not actually I<as> simple as it sounds, particularly with the
89 flexibility POD allows for C<=item>, but it would be good to improve the
90 visual appeal of the HTML generated, and to avoid it having any validation
91 errors. See also L</make HTML install work>, as the layout of installation tree
92 is needed to improve the cross-linking.
94 The addition of C<Pod::Simple> and its related modules may make this task
97 =head2 Make ExtUtils::ParseXS use strict;
99 F<lib/ExtUtils/ParseXS.pm> contains this line
101 # use strict; # One of these days...
103 Simply uncomment it, and fix all the resulting issues :-)
105 The more practical approach, to break the task down into manageable chunks, is
106 to work your way though the code from bottom to top, or if necessary adding
107 extra C<{ ... }> blocks, and turning on strict within them.
109 =head2 Make Schwern poorer
111 We should have tests for everything. When all the core's modules are tested,
112 Schwern has promised to donate to $500 to TPF. We may need volunteers to
113 hold him upside down and shake vigorously in order to actually extract the
116 =head2 Improve the coverage of the core tests
118 Use Devel::Cover to ascertain the core modules' test coverage, then add
119 tests that are currently missing.
123 A full test suite for the B module would be nice.
125 =head2 A decent benchmark
127 C<perlbench> seems impervious to any recent changes made to the perl core. It
128 would be useful to have a reasonable general benchmarking suite that roughly
129 represented what current perl programs do, and measurably reported whether
130 tweaks to the core improve, degrade or don't really affect performance, to
131 guide people attempting to optimise the guts of perl. Gisle would welcome
132 new tests for perlbench.
134 =head2 fix tainting bugs
136 Fix the bugs revealed by running the test suite with the C<-t> switch (via
137 C<make test.taintwarn>).
139 =head2 Dual life everything
141 As part of the "dists" plan, anything that doesn't belong in the smallest perl
142 distribution needs to be dual lifed. Anything else can be too. Figure out what
143 changes would be needed to package that module and its tests up for CPAN, and
144 do so. Test it with older perl releases, and fix the problems you find.
146 To make a minimal perl distribution, it's useful to look at
147 F<t/lib/commonsense.t>.
149 =head2 Move dual-life pod/*.PL into ext
151 Nearly all the dual-life modules have been moved to F<ext>. However, we
152 still need to move F<pod/*.PL> into their respective directories
153 in F<ext/>. They're referenced by (at least) C<plextract> in F<Makefile.SH>
154 and C<utils> in F<win32/Makefile> and F<win32/makefile.ml>, and listed
155 explicitly in F<win32/pod.mak>, F<vms/descrip_mms.template> and F<utils.lst>
157 =head2 POSIX memory footprint
159 Ilya observed that use POSIX; eats memory like there's no tomorrow, and at
160 various times worked to cut it down. There is probably still fat to cut out -
161 for example POSIX passes Exporter some very memory hungry data structures.
163 =head2 embed.pl/makedef.pl
165 There is a script F<embed.pl> that generates several header files to prefix
166 all of Perl's symbols in a consistent way, to provide some semblance of
167 namespace support in C<C>. Functions are declared in F<embed.fnc>, variables
168 in F<interpvar.h>. Quite a few of the functions and variables
169 are conditionally declared there, using C<#ifdef>. However, F<embed.pl>
170 doesn't understand the C macros, so the rules about which symbols are present
171 when is duplicated in F<makedef.pl>. Writing things twice is bad, m'kay.
172 It would be good to teach C<embed.pl> to understand the conditional
173 compilation, and hence remove the duplication, and the mistakes it has caused.
175 =head2 use strict; and AutoLoad
177 Currently if you write
180 use AutoLoader 'AUTOLOAD';
185 print join (' ', No, strict, here), "!\n";
188 then C<use strict;> isn't in force within the autoloaded subroutines. It would
189 be more consistent (and less surprising) to arrange for all lexical pragmas
190 in force at the __END__ block to be in force within each autoloaded subroutine.
192 There's a similar problem with SelfLoader.
194 =head2 profile installman
196 The F<installman> script is slow. All it is doing text processing, which we're
197 told is something Perl is good at. So it would be nice to know what it is doing
198 that is taking so much CPU, and where possible address it.
200 =head2 enable lexical enabling/disabling of inidvidual warnings
202 Currently, warnings can only be enabled or disabled by category. There
203 are times when it would be useful to quash a single warning, not a
206 =head1 Tasks that need a little sysadmin-type knowledge
208 Or if you prefer, tasks that you would learn from, and broaden your skills
211 =head2 make HTML install work
213 There is an C<installhtml> target in the Makefile. It's marked as
214 "experimental". It would be good to get this tested, make it work reliably, and
215 remove the "experimental" tag. This would include
221 Checking that cross linking between various parts of the documentation works.
222 In particular that links work between the modules (files with POD in F<lib/>)
223 and the core documentation (files in F<pod/>)
227 Work out how to split C<perlfunc> into chunks, preferably one per function
228 group, preferably with general case code that could be used elsewhere.
229 Challenges here are correctly identifying the groups of functions that go
230 together, and making the right named external cross-links point to the right
231 page. Things to be aware of are C<-X>, groups such as C<getpwnam> to
232 C<endservent>, two or more C<=items> giving the different parameter lists, such
235 =item substr EXPR,OFFSET,LENGTH,REPLACEMENT
236 =item substr EXPR,OFFSET,LENGTH
237 =item substr EXPR,OFFSET
239 and different parameter lists having different meanings. (eg C<select>)
243 =head2 compressed man pages
245 Be able to install them. This would probably need a configure test to see how
246 the system does compressed man pages (same directory/different directory?
247 same filename/different filename), as well as tweaking the F<installman> script
248 to compress as necessary.
250 =head2 Add a code coverage target to the Makefile
252 Make it easy for anyone to run Devel::Cover on the core's tests. The steps
253 to do this manually are roughly
259 do a normal C<Configure>, but include Devel::Cover as a module to install
260 (see L<INSTALL> for how to do this)
268 cd t; HARNESS_PERL_SWITCHES=-MDevel::Cover ./perl -I../lib harness
272 Process the resulting Devel::Cover database
276 This just give you the coverage of the F<.pm>s. To also get the C level
283 Additionally tell C<Configure> to use the appropriate C compiler flags for
290 (instead of C<make perl>)
294 After running the tests run C<gcov> to generate all the F<.gcov> files.
295 (Including down in the subdirectories of F<ext/>
299 (From the top level perl directory) run C<gcov2perl> on all the C<.gcov> files
300 to get their stats into the cover_db directory.
304 Then process the Devel::Cover database
308 It would be good to add a single switch to C<Configure> to specify that you
309 wanted to perform perl level coverage, and another to specify C level
310 coverage, and have C<Configure> and the F<Makefile> do all the right things
313 =head2 Make Config.pm cope with differences between built and installed perl
315 Quite often vendors ship a perl binary compiled with their (pay-for)
316 compilers. People install a free compiler, such as gcc. To work out how to
317 build extensions, Perl interrogates C<%Config>, so in this situation
318 C<%Config> describes compilers that aren't there, and extension building
319 fails. This forces people into choosing between re-compiling perl themselves
320 using the compiler they have, or only using modules that the vendor ships.
322 It would be good to find a way teach C<Config.pm> about the installation setup,
323 possibly involving probing at install time or later, so that the C<%Config> in
324 a binary distribution better describes the installed machine, when the
325 installed machine differs from the build machine in some significant way.
327 =head2 linker specification files
329 Some platforms mandate that you provide a list of a shared library's external
330 symbols to the linker, so the core already has the infrastructure in place to
331 do this for generating shared perl libraries. My understanding is that the
332 GNU toolchain can accept an optional linker specification file, and restrict
333 visibility just to symbols declared in that file. It would be good to extend
334 F<makedef.pl> to support this format, and to provide a means within
335 C<Configure> to enable it. This would allow Unix users to test that the
336 export list is correct, and to build a perl that does not pollute the global
337 namespace with private symbols, and will fail in the same way as msvc or mingw
338 builds or when using PERL_DL_NONLAZY=1.
340 =head2 Cross-compile support
342 Currently C<Configure> understands C<-Dusecrosscompile> option. This option
343 arranges for building C<miniperl> for TARGET machine, so this C<miniperl> is
344 assumed then to be copied to TARGET machine and used as a replacement of full
347 This could be done little differently. Namely C<miniperl> should be built for
348 HOST and then full C<perl> with extensions should be compiled for TARGET.
349 This, however, might require extra trickery for %Config: we have one config
350 first for HOST and then another for TARGET. Tools like MakeMaker will be
351 mightily confused. Having around two different types of executables and
352 libraries (HOST and TARGET) makes life interesting for Makefiles and
353 shell (and Perl) scripts. There is $Config{run}, normally empty, which
354 can be used as an execution wrapper. Also note that in some
355 cross-compilation/execution environments the HOST and the TARGET do
356 not see the same filesystem(s), the $Config{run} may need to do some
357 file/directory copying back and forth.
361 Make F<pod/roffitall> be updated by F<pod/buildtoc>.
363 =head2 Split "linker" from "compiler"
365 Right now, Configure probes for two commands, and sets two variables:
369 =item * C<cc> (in F<cc.U>)
371 This variable holds the name of a command to execute a C compiler which
372 can resolve multiple global references that happen to have the same
373 name. Usual values are F<cc> and F<gcc>.
374 Fervent ANSI compilers may be called F<c89>. AIX has F<xlc>.
376 =item * C<ld> (in F<dlsrc.U>)
378 This variable indicates the program to be used to link
379 libraries for dynamic loading. On some systems, it is F<ld>.
380 On ELF systems, it should be C<$cc>. Mostly, we'll try to respect
381 the hint file setting.
385 There is an implicit historical assumption from around Perl5.000alpha
386 something, that C<$cc> is also the correct command for linking object files
387 together to make an executable. This may be true on Unix, but it's not true
388 on other platforms, and there are a maze of work arounds in other places (such
389 as F<Makefile.SH>) to cope with this.
391 Ideally, we should create a new variable to hold the name of the executable
392 linker program, probe for it in F<Configure>, and centralise all the special
393 case logic there or in hints files.
395 A small bikeshed issue remains - what to call it, given that C<$ld> is already
396 taken (arguably for the wrong thing now, but on SunOS 4.1 it is the command
397 for creating dynamically-loadable modules) and C<$link> could be confused with
398 the Unix command line executable of the same name, which does something
399 completely different. Andy Dougherty makes the counter argument "In parrot, I
400 tried to call the command used to link object files and libraries into an
401 executable F<link>, since that's what my vaguely-remembered DOS and VMS
402 experience suggested. I don't think any real confusion has ensued, so it's
403 probably a reasonable name for perl5 to use."
405 "Alas, I've always worried that introducing it would make things worse,
406 since now the module building utilities would have to look for
407 C<$Config{link}> and institute a fall-back plan if it weren't found."
408 Although I can see that as confusing, given that C<$Config{d_link}> is true
409 when (hard) links are available.
411 =head2 Configure Windows using PowerShell
413 Currently, Windows uses hard-coded config files based to build the
414 config.h for compiling Perl. Makefiles are also hard-coded and need to be
415 hand edited prior to building Perl. While this makes it easy to create a perl.exe
416 that works across multiple Windows versions, being able to accurately
417 configure a perl.exe for a specific Windows versions and VS C++ would be
418 a nice enhancement. With PowerShell available on Windows XP and up, this
419 may now be possible. Step 1 might be to investigate whether this is possible
420 and use this to clean up our current makefile situation. Step 2 would be to
421 see if there would be a way to use our existing metaconfig units to configure a
422 Windows Perl or whether we go in a separate direction and make it so. Of
423 course, we all know what step 3 is.
425 =head2 decouple -g and -DDEBUGGING
427 Currently F<Configure> automatically adds C<-DDEBUGGING> to the C compiler
428 flags if it spots C<-g> in the optimiser flags. The pre-processor directive
429 C<DEBUGGING> enables F<perl>'s command line C<-D> options, but in the process
430 makes F<perl> slower. It would be good to disentangle this logic, so that
431 C-level debugging with C<-g> and Perl level debugging with C<-D> can easily
432 be enabled independently.
434 =head1 Tasks that need a little C knowledge
436 These tasks would need a little C knowledge, but don't need any specific
437 background or experience with XS, or how the Perl interpreter works
439 =head2 Weed out needless PERL_UNUSED_ARG
441 The C code uses the macro C<PERL_UNUSED_ARG> to stop compilers warning about
442 unused arguments. Often the arguments can't be removed, as there is an
443 external constraint that determines the prototype of the function, so this
444 approach is valid. However, there are some cases where C<PERL_UNUSED_ARG>
445 could be removed. Specifically
451 The prototypes of (nearly all) static functions can be changed
455 Unused arguments generated by short cut macros are wasteful - the short cut
456 macro used can be changed.
460 =head2 Modernize the order of directories in @INC
462 The way @INC is laid out by default, one cannot upgrade core (dual-life)
463 modules without overwriting files. This causes problems for binary
464 package builders. One possible proposal is laid out in this
466 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2002-04/msg02380.html>.
470 Natively 64-bit systems need neither -Duse64bitint nor -Duse64bitall.
471 On these systems, it might be the default compilation mode, and there
472 is currently no guarantee that passing no use64bitall option to the
473 Configure process will build a 32bit perl. Implementing -Duse32bit*
474 options would be nice for perl 5.12.
476 =head2 Profile Perl - am I hot or not?
478 The Perl source code is stable enough that it makes sense to profile it,
479 identify and optimise the hotspots. It would be good to measure the
480 performance of the Perl interpreter using free tools such as cachegrind,
481 gprof, and dtrace, and work to reduce the bottlenecks they reveal.
483 As part of this, the idea of F<pp_hot.c> is that it contains the I<hot> ops,
484 the ops that are most commonly used. The idea is that by grouping them, their
485 object code will be adjacent in the executable, so they have a greater chance
486 of already being in the CPU cache (or swapped in) due to being near another op
489 Except that it's not clear if these really are the most commonly used ops. So
490 as part of exercising your skills with coverage and profiling tools you might
491 want to determine what ops I<really> are the most commonly used. And in turn
492 suggest evictions and promotions to achieve a better F<pp_hot.c>.
494 One piece of Perl code that might make a good testbed is F<installman>.
496 =head2 Allocate OPs from arenas
498 Currently all new OP structures are individually malloc()ed and free()d.
499 All C<malloc> implementations have space overheads, and are now as fast as
500 custom allocates so it would both use less memory and less CPU to allocate
501 the various OP structures from arenas. The SV arena code can probably be
504 Note that Configuring perl with C<-Accflags=-DPL_OP_SLAB_ALLOC> will use
505 Perl_Slab_alloc() to pack optrees into a contiguous block, which is
506 probably superior to the use of OP arenas, esp. from a cache locality
507 standpoint. See L<Profile Perl - am I hot or not?>.
509 =head2 Improve win32/wince.c
511 Currently, numerous functions look virtually, if not completely,
512 identical in both C<win32/wince.c> and C<win32/win32.c> files, which can't
515 =head2 Use secure CRT functions when building with VC8 on Win32
517 Visual C++ 2005 (VC++ 8.x) deprecated a number of CRT functions on the basis
518 that they were "unsafe" and introduced differently named secure versions of
519 them as replacements, e.g. instead of writing
521 FILE* f = fopen(__FILE__, "r");
526 errno_t err = fopen_s(&f, __FILE__, "r");
528 Currently, the warnings about these deprecations have been disabled by adding
529 -D_CRT_SECURE_NO_DEPRECATE to the CFLAGS. It would be nice to remove that
530 warning suppressant and actually make use of the new secure CRT functions.
532 There is also a similar issue with POSIX CRT function names like fileno having
533 been deprecated in favour of ISO C++ conformant names like _fileno. These
534 warnings are also currently suppressed by adding -D_CRT_NONSTDC_NO_DEPRECATE. It
535 might be nice to do as Microsoft suggest here too, although, unlike the secure
536 functions issue, there is presumably little or no benefit in this case.
538 =head2 Fix POSIX::access() and chdir() on Win32
540 These functions currently take no account of DACLs and therefore do not behave
541 correctly in situations where access is restricted by DACLs (as opposed to the
542 read-only attribute).
544 Furthermore, POSIX::access() behaves differently for directories having the
545 read-only attribute set depending on what CRT library is being used. For
546 example, the _access() function in the VC6 and VC7 CRTs (wrongly) claim that
547 such directories are not writable, whereas in fact all directories are writable
548 unless access is denied by DACLs. (In the case of directories, the read-only
549 attribute actually only means that the directory cannot be deleted.) This CRT
550 bug is fixed in the VC8 and VC9 CRTs (but, of course, the directory may still
551 not actually be writable if access is indeed denied by DACLs).
553 For the chdir() issue, see ActiveState bug #74552:
554 http://bugs.activestate.com/show_bug.cgi?id=74552
556 Therefore, DACLs should be checked both for consistency across CRTs and for
559 (Note that perl's -w operator should not be modified to check DACLs. It has
560 been written so that it reflects the state of the read-only attribute, even
561 for directories (whatever CRT is being used), for symmetry with chmod().)
563 =head2 strcat(), strcpy(), strncat(), strncpy(), sprintf(), vsprintf()
565 Maybe create a utility that checks after each libperl.a creation that
566 none of the above (nor sprintf(), vsprintf(), or *SHUDDER* gets())
567 ever creep back to libperl.a.
569 nm libperl.a | ./miniperl -alne '$o = $F[0] if /:$/; print "$o $F[1]" if $F[0] eq "U" && $F[1] =~ /^(?:strn?c(?:at|py)|v?sprintf|gets)$/'
571 Note, of course, that this will only tell whether B<your> platform
572 is using those naughty interfaces.
574 =head2 -D_FORTIFY_SOURCE=2, -fstack-protector
576 Recent glibcs support C<-D_FORTIFY_SOURCE=2> and recent gcc
577 (4.1 onwards?) supports C<-fstack-protector>, both of which give
578 protection against various kinds of buffer overflow problems.
579 These should probably be used for compiling Perl whenever available,
580 Configure and/or hints files should be adjusted to probe for the
581 availability of these features and enable them as appropriate.
583 =head2 Arenas for GPs? For MAGIC?
585 C<struct gp> and C<struct magic> are both currently allocated by C<malloc>.
586 It might be a speed or memory saving to change to using arenas. Or it might
587 not. It would need some suitable benchmarking first. In particular, C<GP>s
588 can probably be changed with minimal compatibility impact (probably nothing
589 outside of the core, or even outside of F<gv.c> allocates them), but they
590 probably aren't allocated/deallocated often enough for a speed saving. Whereas
591 C<MAGIC> is allocated/deallocated more often, but in turn, is also something
592 more externally visible, so changing the rules here may bite external code.
596 Several SV body structs are now the same size, notably PVMG and PVGV, PVAV and
597 PVHV, and PVCV and PVFM. It should be possible to allocate and return same
598 sized bodies from the same actual arena, rather than maintaining one arena for
599 each. This could save 4-6K per thread, of memory no longer tied up in the
600 not-yet-allocated part of an arena.
603 =head1 Tasks that need a knowledge of XS
605 These tasks would need C knowledge, and roughly the level of knowledge of
606 the perl API that comes from writing modules that use XS to interface to
609 =head2 Write an XS cookbook
611 Create pod/perlxscookbook.pod with short, task-focused 'recipes' in XS that
612 demonstrate common tasks and good practices. (Some of these might be
613 extracted from perlguts.) The target audience should be XS novices, who need
614 more examples than perlguts but something less overwhelming than perlapi.
615 Recipes should provide "one pretty good way to do it" instead of TIMTOWTDI.
617 Rather than focusing on interfacing Perl to C libraries, such a cookbook
618 should probably focus on how to optimize Perl routines by re-writing them
619 in XS. This will likely be more motivating to those who mostly work in
620 Perl but are looking to take the next step into XS.
622 Deconstructing and explaining some simpler XS modules could be one way to
623 bootstrap a cookbook. (List::Util? Class::XSAccessor? Tree::Ternary_XS?)
624 Another option could be deconstructing the implementation of some simpler
627 =head2 Allow XSUBs to inline themselves as OPs
629 For a simple XSUB, often the subroutine dispatch takes more time than the
630 XSUB itself. The tokeniser already has the ability to inline constant
631 subroutines - it would be good to provide a way to inline other subroutines.
633 Specifically, simplest approach looks to be to allow an XSUB to provide an
634 alternative implementation of itself as a custom OP. A new flag bit in
635 C<CvFLAGS()> would signal to the peephole optimiser to take an optree
638 b <@> leave[1 ref] vKP/REFC ->(end)
640 2 <;> nextstate(main 1 -e:1) v:{ ->3
641 a <2> sassign vKS/2 ->b
642 8 <1> entersub[t2] sKS/TARG,1 ->9
645 4 <$> const(IV 1) sM ->5
646 6 <1> rv2av[t1] lKM/1 ->7
648 - <1> ex-rv2cv sK ->-
649 7 <$> gv(*x) s/EARLYCV ->8
650 - <1> ex-rv2sv sKRM*/1 ->a
653 perform the symbol table lookup of C<rv2cv> and C<gv(*x)>, locate the
654 pointer to the custom OP that provides the direct implementation, and re-
655 write the optree something like:
657 b <@> leave[1 ref] vKP/REFC ->(end)
659 2 <;> nextstate(main 1 -e:1) v:{ ->3
660 a <2> sassign vKS/2 ->b
664 4 <$> const(IV 1) sM ->5
665 6 <1> rv2av[t1] lKM/1 ->7
667 - <1> ex-rv2cv sK ->-
668 - <$> ex-gv(*x) s/EARLYCV ->7
669 - <1> ex-rv2sv sKRM*/1 ->a
672 I<i.e.> the C<gv(*)> OP has been nulled and spliced out of the execution
673 path, and the C<entersub> OP has been replaced by the custom op.
675 This approach should provide a measurable speed up to simple XSUBs inside
676 tight loops. Initially one would have to write the OP alternative
677 implementation by hand, but it's likely that this should be reasonably
678 straightforward for the type of XSUB that would benefit the most. Longer
679 term, once the run-time implementation is proven, it should be possible to
680 progressively update ExtUtils::ParseXS to generate OP implementations for
683 =head2 Remove the use of SVs as temporaries in dump.c
685 F<dump.c> contains debugging routines to dump out the contains of perl data
686 structures, such as C<SV>s, C<AV>s and C<HV>s. Currently, the dumping code
687 B<uses> C<SV>s for its temporary buffers, which was a logical initial
688 implementation choice, as they provide ready made memory handling.
690 However, they also lead to a lot of confusion when it happens that what you're
691 trying to debug is seen by the code in F<dump.c>, correctly or incorrectly, as
692 a temporary scalar it can use for a temporary buffer. It's also not possible
693 to dump scalars before the interpreter is properly set up, such as during
694 ithreads cloning. It would be good to progressively replace the use of scalars
695 as string accumulation buffers with something much simpler, directly allocated
696 by C<malloc>. The F<dump.c> code is (or should be) only producing 7 bit
697 US-ASCII, so output character sets are not an issue.
699 Producing and proving an internal simple buffer allocation would make it easier
700 to re-write the internals of the PerlIO subsystem to avoid using C<SV>s for
701 B<its> buffers, use of which can cause problems similar to those of F<dump.c>,
704 =head2 safely supporting POSIX SA_SIGINFO
706 Some years ago Jarkko supplied patches to provide support for the POSIX
707 SA_SIGINFO feature in Perl, passing the extra data to the Perl signal handler.
709 Unfortunately, it only works with "unsafe" signals, because under safe
710 signals, by the time Perl gets to run the signal handler, the extra
711 information has been lost. Moreover, it's not easy to store it somewhere,
712 as you can't call mutexs, or do anything else fancy, from inside a signal
715 So it strikes me that we could provide safe SA_SIGINFO support
721 Provide global variables for two file descriptors
725 When the first request is made via C<sigaction> for C<SA_SIGINFO>, create a
726 pipe, store the reader in one, the writer in the other
730 In the "safe" signal handler (C<Perl_csighandler()>/C<S_raise_signal()>), if
731 the C<siginfo_t> pointer non-C<NULL>, and the writer file handle is open,
737 serialise signal number, C<struct siginfo_t> (or at least the parts we care
738 about) into a small auto char buff
742 C<write()> that (non-blocking) to the writer fd
748 if it writes 100%, flag the signal in a counter of "signals on the pipe" akin
749 to the current per-signal-number counts
753 if it writes 0%, assume the pipe is full. Flag the data as lost?
757 if it writes partially, croak a panic, as your OS is broken.
765 in the regular C<PERL_ASYNC_CHECK()> processing, if there are "signals on
766 the pipe", read the data out, deserialise, build the Perl structures on
767 the stack (code in C<Perl_sighandler()>, the "unsafe" handler), and call as
772 I think that this gets us decent C<SA_SIGINFO> support, without the current risk
773 of running Perl code inside the signal handler context. (With all the dangers
774 of things like C<malloc> corruption that that currently offers us)
776 For more information see the thread starting with this message:
777 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-03/msg00305.html
779 =head2 autovivification
781 Make all autovivification consistent w.r.t LVALUE/RVALUE and strict/no strict;
783 This task is incremental - even a little bit of work on it will help.
785 =head2 Unicode in Filenames
787 chdir, chmod, chown, chroot, exec, glob, link, lstat, mkdir, open,
788 opendir, qx, readdir, readlink, rename, rmdir, stat, symlink, sysopen,
789 system, truncate, unlink, utime, -X. All these could potentially accept
790 Unicode filenames either as input or output (and in the case of system
791 and qx Unicode in general, as input or output to/from the shell).
792 Whether a filesystem - an operating system pair understands Unicode in
795 Known combinations that have some level of understanding include
796 Microsoft NTFS, Apple HFS+ (In Mac OS 9 and X) and Apple UFS (in Mac
797 OS X), NFS v4 is rumored to be Unicode, and of course Plan 9. How to
798 create Unicode filenames, what forms of Unicode are accepted and used
799 (UCS-2, UTF-16, UTF-8), what (if any) is the normalization form used,
800 and so on, varies. Finding the right level of interfacing to Perl
801 requires some thought. Remember that an OS does not implicate a
804 (The Windows -C command flag "wide API support" has been at least
805 temporarily retired in 5.8.1, and the -C has been repurposed, see
808 Most probably the right way to do this would be this:
809 L</"Virtualize operating system access">.
811 =head2 Unicode in %ENV
813 Currently the %ENV entries are always byte strings.
814 See L</"Virtualize operating system access">.
816 =head2 Unicode and glob()
818 Currently glob patterns and filenames returned from File::Glob::glob()
819 are always byte strings. See L</"Virtualize operating system access">.
821 =head2 use less 'memory'
823 Investigate trade offs to switch out perl's choices on memory usage.
824 Particularly perl should be able to give memory back.
826 This task is incremental - even a little bit of work on it will help.
828 =head2 Re-implement C<:unique> in a way that is actually thread-safe
830 The old implementation made bad assumptions on several levels. A good 90%
831 solution might be just to make C<:unique> work to share the string buffer
832 of SvPVs. That way large constant strings can be shared between ithreads,
833 such as the configuration information in F<Config>.
835 =head2 Make tainting consistent
837 Tainting would be easier to use if it didn't take documented shortcuts and
838 allow taint to "leak" everywhere within an expression.
840 =head2 readpipe(LIST)
842 system() accepts a LIST syntax (and a PROGRAM LIST syntax) to avoid
843 running a shell. readpipe() (the function behind qx//) could be similarly
846 =head2 Audit the code for destruction ordering assumptions
850 /* Need to check SvMAGICAL, as during global destruction it may be that
851 AvARYLEN(av) has been freed before av, and hence the SvANY() pointer
852 is now part of the linked list of SV heads, rather than pointing to
853 the original body. */
854 /* FIXME - audit the code for other bugs like this one. */
856 adding the C<SvMAGICAL> check to
858 if (AvARYLEN(av) && SvMAGICAL(AvARYLEN(av))) {
859 MAGIC *mg = mg_find (AvARYLEN(av), PERL_MAGIC_arylen);
861 Go through the core and look for similar assumptions that SVs have particular
862 types, as all bets are off during global destruction.
864 =head2 Extend PerlIO and PerlIO::Scalar
866 PerlIO::Scalar doesn't know how to truncate(). Implementing this
867 would require extending the PerlIO vtable.
869 Similarly the PerlIO vtable doesn't know about formats (write()), or
870 about stat(), or chmod()/chown(), utime(), or flock().
872 (For PerlIO::Scalar it's hard to see what e.g. mode bits or ownership
875 PerlIO doesn't do directories or symlinks, either: mkdir(), rmdir(),
876 opendir(), closedir(), seekdir(), rewinddir(), glob(); symlink(),
879 See also L</"Virtualize operating system access">.
881 =head2 -C on the #! line
883 It should be possible to make -C work correctly if found on the #! line,
884 given that all perl command line options are strict ASCII, and -C changes
885 only the interpretation of non-ASCII characters, and not for the script file
886 handle. To make it work needs some investigation of the ordering of function
887 calls during startup, and (by implication) a bit of tweaking of that order.
889 =head2 Organize error messages
891 Perl's diagnostics (error messages, see L<perldiag>) could use
892 reorganizing and formalizing so that each error message has its
893 stable-for-all-eternity unique id, categorized by severity, type, and
894 subsystem. (The error messages would be listed in a datafile outside
895 of the Perl source code, and the source code would only refer to the
896 messages by the id.) This clean-up and regularizing should apply
897 for all croak() messages.
899 This would enable all sorts of things: easier translation/localization
900 of the messages (though please do keep in mind the caveats of
901 L<Locale::Maketext> about too straightforward approaches to
902 translation), filtering by severity, and instead of grepping for a
903 particular error message one could look for a stable error id. (Of
904 course, changing the error messages by default would break all the
905 existing software depending on some particular error message...)
907 This kind of functionality is known as I<message catalogs>. Look for
908 inspiration for example in the catgets() system, possibly even use it
909 if available-- but B<only> if available, all platforms will B<not>
912 For the really pure at heart, consider extending this item to cover
913 also the warning messages (see L<perllexwarn>, C<warnings.pl>).
915 =head1 Tasks that need a knowledge of the interpreter
917 These tasks would need C knowledge, and knowledge of how the interpreter works,
918 or a willingness to learn.
920 =head2 forbid labels with keyword names
922 Currently C<goto keyword> "computes" the label value:
924 $ perl -e 'goto print'
925 Can't find label 1 at -e line 1.
927 It is controversial if the right way to avoid the confusion is to forbid
928 labels with keyword names, or if it would be better to always treat
929 bareword expressions after a "goto" as a label and never as a keyword.
931 =head2 truncate() prototype
933 The prototype of truncate() is currently C<$$>. It should probably
934 be C<*$> instead. (This is changed in F<opcode.pl>)
936 =head2 decapsulation of smart match argument
938 Currently C<$foo ~~ $object> will die with the message "Smart matching a
939 non-overloaded object breaks encapsulation". It would be nice to allow
940 to bypass this by using explictly the syntax C<$foo ~~ %$object> or
943 =head2 error reporting of [$a ; $b]
945 Using C<;> inside brackets is a syntax error, and we don't propose to change
946 that by giving it any meaning. However, it's not reported very helpfully:
948 $ perl -e '$a = [$b; $c];'
949 syntax error at -e line 1, near "$b;"
950 syntax error at -e line 1, near "$c]"
951 Execution of -e aborted due to compilation errors.
953 It should be possible to hook into the tokeniser or the lexer, so that when a
954 C<;> is parsed where it is not legal as a statement terminator (ie inside
955 C<{}> used as a hashref, C<[]> or C<()>) it issues an error something like
956 I<';' isn't legal inside an expression - if you need multiple statements use a
957 do {...} block>. See the thread starting at
958 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-09/msg00573.html
960 =head2 lexicals used only once
964 $ perl -we '$pie = 42'
965 Name "main::pie" used only once: possible typo at -e line 1.
969 $ perl -we 'my $pie = 42'
971 Logically all lexicals used only once should warn, if the user asks for
972 warnings. An unworked RT ticket (#5087) has been open for almost seven
973 years for this discrepancy.
977 The handling of Unicode is unclean in many places. For example, the regexp
978 engine matches in Unicode semantics whenever the string or the pattern is
979 flagged as UTF-8, but that should not be dependent on an internal storage
980 detail of the string.
982 =head2 Properly Unicode safe tokeniser and pads.
984 The tokeniser isn't actually very UTF-8 clean. C<use utf8;> is a hack -
985 variable names are stored in stashes as raw bytes, without the utf-8 flag
986 set. The pad API only takes a C<char *> pointer, so that's all bytes too. The
987 tokeniser ignores the UTF-8-ness of C<PL_rsfp>, or any SVs returned from
988 source filters. All this could be fixed.
990 =head2 state variable initialization in list context
992 Currently this is illegal:
994 state ($a, $b) = foo();
996 In Perl 6, C<state ($a) = foo();> and C<(state $a) = foo();> have different
997 semantics, which is tricky to implement in Perl 5 as currently they produce
998 the same opcode trees. The Perl 6 design is firm, so it would be good to
999 implement the necessary code in Perl 5. There are comments in
1000 C<Perl_newASSIGNOP()> that show the code paths taken by various assignment
1001 constructions involving state variables.
1003 =head2 Implement $value ~~ 0 .. $range
1005 It would be nice to extend the syntax of the C<~~> operator to also
1006 understand numeric (and maybe alphanumeric) ranges.
1008 =head2 A does() built-in
1010 Like ref(), only useful. It would call the C<DOES> method on objects; it
1011 would also tell whether something can be dereferenced as an
1012 array/hash/etc., or used as a regexp, etc.
1013 L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-03/msg00481.html>
1015 =head2 Tied filehandles and write() don't mix
1017 There is no method on tied filehandles to allow them to be called back by
1020 =head2 Propagate compilation hints to the debugger
1022 Currently a debugger started with -dE on the command-line doesn't see the
1023 features enabled by -E. More generally hints (C<$^H> and C<%^H>) aren't
1024 propagated to the debugger. Probably it would be a good thing to propagate
1025 hints from the innermost non-C<DB::> scope: this would make code eval'ed
1026 in the debugger see the features (and strictures, etc.) currently in
1029 =head2 Attach/detach debugger from running program
1031 The old perltodo notes "With C<gdb>, you can attach the debugger to a running
1032 program if you pass the process ID. It would be good to do this with the Perl
1033 debugger on a running Perl program, although I'm not sure how it would be
1034 done." ssh and screen do this with named pipes in /tmp. Maybe we can too.
1036 =head2 LVALUE functions for lists
1038 The old perltodo notes that lvalue functions don't work for list or hash
1039 slices. This would be good to fix.
1041 =head2 regexp optimiser optional
1043 The regexp optimiser is not optional. It should configurable to be, to allow
1044 its performance to be measured, and its bugs to be easily demonstrated.
1046 =head2 C</w> regex modifier
1048 That flag would enable to match whole words, and also to interpolate
1049 arrays as alternations. With it, C</P/w> would be roughly equivalent to:
1051 do { local $"='|'; /\b(?:P)\b/ }
1053 See L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-01/msg00400.html>
1056 =head2 optional optimizer
1058 Make the peephole optimizer optional. Currently it performs two tasks as
1059 it walks the optree - genuine peephole optimisations, and necessary fixups of
1060 ops. It would be good to find an efficient way to switch out the
1061 optimisations whilst keeping the fixups.
1063 =head2 You WANT *how* many
1065 Currently contexts are void, scalar and list. split has a special mechanism in
1066 place to pass in the number of return values wanted. It would be useful to
1067 have a general mechanism for this, backwards compatible and little speed hit.
1068 This would allow proposals such as short circuiting sort to be implemented
1069 as a module on CPAN.
1071 =head2 lexical aliases
1073 Allow lexical aliases (maybe via the syntax C<my \$alias = \$foo>.
1075 =head2 entersub XS vs Perl
1077 At the moment pp_entersub is huge, and has code to deal with entering both
1078 perl and XS subroutines. Subroutine implementations rarely change between
1079 perl and XS at run time, so investigate using 2 ops to enter subs (one for
1080 XS, one for perl) and swap between if a sub is redefined.
1084 Self-ties are currently illegal because they caused too many segfaults. Maybe
1085 the causes of these could be tracked down and self-ties on all types
1088 =head2 Optimize away @_
1090 The old perltodo notes "Look at the "reification" code in C<av.c>".
1092 =head2 Virtualize operating system access
1094 Implement a set of "vtables" that virtualizes operating system access
1095 (open(), mkdir(), unlink(), readdir(), getenv(), etc.) At the very
1096 least these interfaces should take SVs as "name" arguments instead of
1097 bare char pointers; probably the most flexible and extensible way
1098 would be for the Perl-facing interfaces to accept HVs. The system
1099 needs to be per-operating-system and per-file-system
1100 hookable/filterable, preferably both from XS and Perl level
1101 (L<perlport/"Files and Filesystems"> is good reading at this point,
1102 in fact, all of L<perlport> is.)
1104 This has actually already been implemented (but only for Win32),
1105 take a look at F<iperlsys.h> and F<win32/perlhost.h>. While all Win32
1106 variants go through a set of "vtables" for operating system access,
1107 non-Win32 systems currently go straight for the POSIX/Unix-style
1108 system/library call. Similar system as for Win32 should be
1109 implemented for all platforms. The existing Win32 implementation
1110 probably does not need to survive alongside this proposed new
1111 implementation, the approaches could be merged.
1113 What would this give us? One often-asked-for feature this would
1114 enable is using Unicode for filenames, and other "names" like %ENV,
1115 usernames, hostnames, and so forth.
1116 (See L<perlunicode/"When Unicode Does Not Happen">.)
1118 But this kind of virtualization would also allow for things like
1119 virtual filesystems, virtual networks, and "sandboxes" (though as long
1120 as dynamic loading of random object code is allowed, not very safe
1121 sandboxes since external code of course know not of Perl's vtables).
1122 An example of a smaller "sandbox" is that this feature can be used to
1123 implement per-thread working directories: Win32 already does this.
1125 See also L</"Extend PerlIO and PerlIO::Scalar">.
1127 =head2 Investigate PADTMP hash pessimisation
1129 The peephole optimiser converts constants used for hash key lookups to shared
1130 hash key scalars. Under ithreads, something is undoing this work.
1131 See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-09/msg00793.html
1133 =head2 Store the current pad in the OP slab allocator
1136 I hope that I got that "current pad" part correct
1138 Currently we leak ops in various cases of parse failure. I suggested that we
1139 could solve this by always using the op slab allocator, and walking it to
1140 free ops. Dave comments that as some ops are already freed during optree
1141 creation one would have to mark which ops are freed, and not double free them
1142 when walking the slab. He notes that one problem with this is that for some ops
1143 you have to know which pad was current at the time of allocation, which does
1144 change. I suggested storing a pointer to the current pad in the memory allocated
1145 for the slab, and swapping to a new slab each time the pad changes. Dave thinks
1146 that this would work.
1148 =head2 repack the optree
1150 Repacking the optree after execution order is determined could allow
1151 removal of NULL ops, and optimal ordering of OPs with respect to cache-line
1152 filling. The slab allocator could be reused for this purpose. I think that
1153 the best way to do this is to make it an optional step just before the
1154 completed optree is attached to anything else, and to use the slab allocator
1155 unchanged, so that freeing ops is identical whether or not this step runs.
1156 Note that the slab allocator allocates ops downwards in memory, so one would
1157 have to actually "allocate" the ops in reverse-execution order to get them
1158 contiguous in memory in execution order.
1160 See http://www.nntp.perl.org/group/perl.perl5.porters/2007/12/msg131975.html
1162 Note that running this copy, and then freeing all the old location ops would
1163 cause their slabs to be freed, which would eliminate possible memory wastage if
1164 the previous suggestion is implemented, and we swap slabs more frequently.
1166 =head2 eliminate incorrect line numbers in warnings
1174 } elsif ($undef == 0) {
1177 used to produce this output:
1179 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1180 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1182 where the line of the second warning was misreported - it should be line 5.
1183 Rafael fixed this - the problem arose because there was no nextstate OP
1184 between the execution of the C<if> and the C<elsif>, hence C<PL_curcop> still
1185 reports that the currently executing line is line 4. The solution was to inject
1186 a nextstate OPs for each C<elsif>, although it turned out that the nextstate
1187 OP needed to be a nulled OP, rather than a live nextstate OP, else other line
1188 numbers became misreported. (Jenga!)
1190 The problem is more general than C<elsif> (although the C<elsif> case is the
1191 most common and the most confusing). Ideally this code
1201 would produce this output
1203 Use of uninitialized value $undef in addition (+) at wrong.pl line 4.
1204 Use of uninitialized value $undef in addition (+) at wrong.pl line 7.
1206 (rather than lines 4 and 5), but this would seem to require every OP to carry
1207 (at least) line number information.
1209 What might work is to have an optional line number in memory just before the
1210 BASEOP structure, with a flag bit in the op to say whether it's present.
1211 Initially during compile every OP would carry its line number. Then add a late
1212 pass to the optimiser (potentially combined with L</repack the optree>) which
1213 looks at the two ops on every edge of the graph of the execution path. If
1214 the line number changes, flags the destination OP with this information.
1215 Once all paths are traced, replace every op with the flag with a
1216 nextstate-light op (that just updates C<PL_curcop>), which in turn then passes
1217 control on to the true op. All ops would then be replaced by variants that
1218 do not store the line number. (Which, logically, why it would work best in
1219 conjunction with L</repack the optree>, as that is already copying/reallocating
1222 (Although I should note that we're not certain that doing this for the general
1225 =head2 optimize tail-calls
1227 Tail-calls present an opportunity for broadly applicable optimization;
1228 anywhere that C<< return foo(...) >> is called, the outer return can
1229 be replaced by a goto, and foo will return directly to the outer
1230 caller, saving (conservatively) 25% of perl's call&return cost, which
1231 is relatively higher than in C. The scheme language is known to do
1232 this heavily. B::Concise provides good insight into where this
1233 optimization is possible, ie anywhere entersub,leavesub op-sequence
1236 perl -MO=Concise,-exec,a,b,-main -e 'sub a{ 1 }; sub b {a()}; b(2)'
1238 Bottom line on this is probably a new pp_tailcall function which
1239 combines the code in pp_entersub, pp_leavesub. This should probably
1240 be done 1st in XS, and using B::Generate to patch the new OP into the
1243 =head2 Add C<00dddd>
1245 It has been proposed that octal constants be specifiable through the syntax
1246 C<0oddddd>, parallel to the existing construct to specify hex constants
1251 Tasks that will get your name mentioned in the description of the "Highlights
1254 =head2 make ithreads more robust
1256 Generally make ithreads more robust. See also L</iCOW>
1258 This task is incremental - even a little bit of work on it will help, and
1259 will be greatly appreciated.
1261 One bit would be to write the missing code in sv.c:Perl_dirp_dup.
1263 Fix Perl_sv_dup, et al so that threads can return objects.
1267 Sarathy and Arthur have a proposal for an improved Copy On Write which
1268 specifically will be able to COW new ithreads. If this can be implemented
1269 it would be a good thing.
1271 =head2 (?{...}) closures in regexps
1273 Fix (or rewrite) the implementation of the C</(?{...})/> closures.
1275 =head2 A re-entrant regexp engine
1277 This will allow the use of a regex from inside (?{ }), (??{ }) and
1278 (?(?{ })|) constructs.
1280 =head2 Add class set operations to regexp engine
1282 Apparently these are quite useful. Anyway, Jeffery Friedl wants them.
1284 demerphq has this on his todo list, but right at the bottom.
1287 =head1 Tasks for microperl
1290 [ Each and every one of these may be obsolete, but they were listed
1291 in the old Todo.micro file]
1294 =head2 make creating uconfig.sh automatic
1296 =head2 make creating Makefile.micro automatic
1298 =head2 do away with fork/exec/wait?
1300 (system, popen should be enough?)
1302 =head2 some of the uconfig.sh really needs to be probed (using cc) in buildtime:
1304 (uConfigure? :-) native datatype widths and endianness come to mind