note strict/lax version requirements in documentation
[p5sagit/p5-mst-13.2.git] / pod / perltodo.pod
CommitLineData
7711098a 1=head1 NAME
2
3perltodo - Perl TO-DO List
4
5=head1 DESCRIPTION
e50bb9a1 6
049aabcb 7This is a list of wishes for Perl. The most up to date version of this file
8is at http://perl5.git.perl.org/perl.git/blob_plain/HEAD:/pod/perltodo.pod
9
10The tasks we think are smaller or easier are listed first. Anyone is welcome
11to work on any of these, but it's a good idea to first contact
12I<perl5-porters@perl.org> to avoid duplication of effort, and to learn from
13any previous attempts. By all means contact a pumpking privately first if you
14prefer.
e50bb9a1 15
0bdfc961 16Whilst patches to make the list shorter are most welcome, ideas to add to
17the list are also encouraged. Check the perl5-porters archives for past
18ideas, and any discussion about them. One set of archives may be found at:
e50bb9a1 19
0bdfc961 20 http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/
938c8732 21
617eabfa 22What can we offer you in return? Fame, fortune, and everlasting glory? Maybe
23not, but if your patch is incorporated, then we'll add your name to the
24F<AUTHORS> file, which ships in the official distribution. How many other
25programming languages offer you 1 line of immortality?
938c8732 26
0bdfc961 27=head1 Tasks that only need Perl knowledge
e50bb9a1 28
8401756a 29=head2 Improve Porting/cmpVERSION.pl to work from git tags
30
31See F<Porting/release_managers_guide.pod> for a bit more detail.
32
de2b17d8 33=head2 Migrate t/ from custom TAP generation
34
35Many tests below F<t/> still generate TAP by "hand", rather than using library
36functions. As explained in L<perlhack/Writing a test>, tests in F<t/> are
37written in a particular way to test that more complex constructions actually
38work before using them routinely. Hence they don't use C<Test::More>, but
39instead there is an intentionally simpler library, F<t/test.pl>. However,
40quite a few tests in F<t/> have not been refactored to use it. Refactoring
41any of these tests, one at a time, is a useful thing TODO.
42
0d8e5a42 43The subdirectories F<base>, F<cmd> and F<comp>, that contain the most
44basic tests, should be excluded from this task.
45
08063926 46=head2 Test that regen.pl was run
47
48There are various generated files shipped with the perl distribution, for
49things like header files generate from data. The generation scripts are
50written in perl, and all can be run by F<regen.pl>. However, because they're
51written in perl, we can't run them before we've built perl. We can't run them
52as part of the F<Makefile>, because changing files underneath F<make> confuses
53it completely, and we don't want to run them automatically anyway, as they
54change files shipped by the distribution, something we seek not do to.
55
56If someone changes the data, but forgets to re-run F<regen.pl> then the
57generated files are out of sync. It would be good to have a test in
58F<t/porting> that checks that the generated files are in sync, and fails
59otherwise, to alert someone before they make a poor commit. I suspect that this
60would require adapting the scripts run from F<regen.pl> to have dry-run
61options, and invoking them with these, or by refactoring them into a library
62that does the generation, which can be called by the scripts, and by the test.
63
0be987a2 64=head2 Automate perldelta generation
65
66The perldelta file accompanying each release summaries the major changes.
67It's mostly manually generated currently, but some of that could be
68automated with a bit of perl, specifically the generation of
69
70=over
71
72=item Modules and Pragmata
73
74=item New Documentation
75
76=item New Tests
77
78=back
79
80See F<Porting/how_to_write_a_perldelta.pod> for details.
81
5a176cbc 82=head2 Remove duplication of test setup.
83
84Schwern notes, that there's duplication of code - lots and lots of tests have
85some variation on the big block of C<$Is_Foo> checks. We can safely put this
86into a file, change it to build an C<%Is> hash and require it. Maybe just put
87it into F<test.pl>. Throw in the handy tainting subroutines.
88
87a942b1 89=head2 POD -E<gt> HTML conversion in the core still sucks
e50bb9a1 90
938c8732 91Which is crazy given just how simple POD purports to be, and how simple HTML
adebf063 92can be. It's not actually I<as> simple as it sounds, particularly with the
93flexibility POD allows for C<=item>, but it would be good to improve the
94visual appeal of the HTML generated, and to avoid it having any validation
95errors. See also L</make HTML install work>, as the layout of installation tree
96is needed to improve the cross-linking.
938c8732 97
dc0fb092 98The addition of C<Pod::Simple> and its related modules may make this task
99easier to complete.
100
0befdfba 101=head2 Make ExtUtils::ParseXS use strict;
102
103F<lib/ExtUtils/ParseXS.pm> contains this line
104
105 # use strict; # One of these days...
106
107Simply uncomment it, and fix all the resulting issues :-)
108
109The more practical approach, to break the task down into manageable chunks, is
110to work your way though the code from bottom to top, or if necessary adding
111extra C<{ ... }> blocks, and turning on strict within them.
112
0bdfc961 113=head2 Make Schwern poorer
e50bb9a1 114
613bd4f7 115We should have tests for everything. When all the core's modules are tested,
0bdfc961 116Schwern has promised to donate to $500 to TPF. We may need volunteers to
117hold him upside down and shake vigorously in order to actually extract the
118cash.
3958b146 119
0bdfc961 120=head2 Improve the coverage of the core tests
e50bb9a1 121
e1020413 122Use Devel::Cover to ascertain the core modules' test coverage, then add
02f21748 123tests that are currently missing.
30222c0f 124
0bdfc961 125=head2 test B
e50bb9a1 126
0bdfc961 127A full test suite for the B module would be nice.
e50bb9a1 128
0bdfc961 129=head2 A decent benchmark
e50bb9a1 130
617eabfa 131C<perlbench> seems impervious to any recent changes made to the perl core. It
0bdfc961 132would be useful to have a reasonable general benchmarking suite that roughly
133represented what current perl programs do, and measurably reported whether
134tweaks to the core improve, degrade or don't really affect performance, to
135guide people attempting to optimise the guts of perl. Gisle would welcome
136new tests for perlbench.
6168cf99 137
0bdfc961 138=head2 fix tainting bugs
6168cf99 139
0bdfc961 140Fix the bugs revealed by running the test suite with the C<-t> switch (via
141C<make test.taintwarn>).
e50bb9a1 142
0bdfc961 143=head2 Dual life everything
e50bb9a1 144
0bdfc961 145As part of the "dists" plan, anything that doesn't belong in the smallest perl
146distribution needs to be dual lifed. Anything else can be too. Figure out what
147changes would be needed to package that module and its tests up for CPAN, and
148do so. Test it with older perl releases, and fix the problems you find.
e50bb9a1 149
a393eb28 150To make a minimal perl distribution, it's useful to look at
151F<t/lib/commonsense.t>.
152
dfb56e28 153=head2 Move dual-life pod/*.PL into ext
c2aba5b8 154
dfb56e28 155Nearly all the dual-life modules have been moved to F<ext>. However, we
156still need to move F<pod/*.PL> into their respective directories
764e6bc7 157in F<ext/>. They're referenced by (at least) C<plextract> in F<Makefile.SH>
158and C<utils> in F<win32/Makefile> and F<win32/makefile.ml>, and listed
159explicitly in F<win32/pod.mak>, F<vms/descrip_mms.template> and F<utils.lst>
160
0bdfc961 161=head2 POSIX memory footprint
e50bb9a1 162
0bdfc961 163Ilya observed that use POSIX; eats memory like there's no tomorrow, and at
164various times worked to cut it down. There is probably still fat to cut out -
165for example POSIX passes Exporter some very memory hungry data structures.
e50bb9a1 166
eed36644 167=head2 embed.pl/makedef.pl
168
169There is a script F<embed.pl> that generates several header files to prefix
170all of Perl's symbols in a consistent way, to provide some semblance of
171namespace support in C<C>. Functions are declared in F<embed.fnc>, variables
907b3e23 172in F<interpvar.h>. Quite a few of the functions and variables
eed36644 173are conditionally declared there, using C<#ifdef>. However, F<embed.pl>
174doesn't understand the C macros, so the rules about which symbols are present
175when is duplicated in F<makedef.pl>. Writing things twice is bad, m'kay.
176It would be good to teach C<embed.pl> to understand the conditional
177compilation, and hence remove the duplication, and the mistakes it has caused.
e50bb9a1 178
801de10e 179=head2 use strict; and AutoLoad
180
181Currently if you write
182
183 package Whack;
184 use AutoLoader 'AUTOLOAD';
185 use strict;
186 1;
187 __END__
188 sub bloop {
189 print join (' ', No, strict, here), "!\n";
190 }
191
192then C<use strict;> isn't in force within the autoloaded subroutines. It would
193be more consistent (and less surprising) to arrange for all lexical pragmas
194in force at the __END__ block to be in force within each autoloaded subroutine.
195
773b3597 196There's a similar problem with SelfLoader.
197
91d0cbf6 198=head2 profile installman
199
200The F<installman> script is slow. All it is doing text processing, which we're
201told is something Perl is good at. So it would be nice to know what it is doing
202that is taking so much CPU, and where possible address it.
203
a9ed9b74 204=head2 enable lexical enabling/disabling of inidvidual warnings
205
206Currently, warnings can only be enabled or disabled by category. There
207are times when it would be useful to quash a single warning, not a
208whole category.
91d0cbf6 209
0bdfc961 210=head1 Tasks that need a little sysadmin-type knowledge
e50bb9a1 211
0bdfc961 212Or if you prefer, tasks that you would learn from, and broaden your skills
213base...
e50bb9a1 214
cd793d32 215=head2 make HTML install work
e50bb9a1 216
adebf063 217There is an C<installhtml> target in the Makefile. It's marked as
218"experimental". It would be good to get this tested, make it work reliably, and
219remove the "experimental" tag. This would include
220
221=over 4
222
223=item 1
224
225Checking that cross linking between various parts of the documentation works.
226In particular that links work between the modules (files with POD in F<lib/>)
227and the core documentation (files in F<pod/>)
228
229=item 2
230
617eabfa 231Work out how to split C<perlfunc> into chunks, preferably one per function
232group, preferably with general case code that could be used elsewhere.
233Challenges here are correctly identifying the groups of functions that go
234together, and making the right named external cross-links point to the right
235page. Things to be aware of are C<-X>, groups such as C<getpwnam> to
236C<endservent>, two or more C<=items> giving the different parameter lists, such
237as
adebf063 238
239 =item substr EXPR,OFFSET,LENGTH,REPLACEMENT
adebf063 240 =item substr EXPR,OFFSET,LENGTH
adebf063 241 =item substr EXPR,OFFSET
242
243and different parameter lists having different meanings. (eg C<select>)
244
245=back
3a89a73c 246
0bdfc961 247=head2 compressed man pages
248
249Be able to install them. This would probably need a configure test to see how
250the system does compressed man pages (same directory/different directory?
251same filename/different filename), as well as tweaking the F<installman> script
252to compress as necessary.
253
30222c0f 254=head2 Add a code coverage target to the Makefile
255
256Make it easy for anyone to run Devel::Cover on the core's tests. The steps
257to do this manually are roughly
258
259=over 4
260
261=item *
262
263do a normal C<Configure>, but include Devel::Cover as a module to install
264(see F<INSTALL> for how to do this)
265
266=item *
267
268 make perl
269
270=item *
271
272 cd t; HARNESS_PERL_SWITCHES=-MDevel::Cover ./perl -I../lib harness
273
274=item *
275
276Process the resulting Devel::Cover database
277
278=back
279
280This just give you the coverage of the F<.pm>s. To also get the C level
281coverage you need to
282
283=over 4
284
285=item *
286
287Additionally tell C<Configure> to use the appropriate C compiler flags for
288C<gcov>
289
290=item *
291
292 make perl.gcov
293
294(instead of C<make perl>)
295
296=item *
297
298After running the tests run C<gcov> to generate all the F<.gcov> files.
299(Including down in the subdirectories of F<ext/>
300
301=item *
302
303(From the top level perl directory) run C<gcov2perl> on all the C<.gcov> files
304to get their stats into the cover_db directory.
305
306=item *
307
308Then process the Devel::Cover database
309
310=back
311
312It would be good to add a single switch to C<Configure> to specify that you
313wanted to perform perl level coverage, and another to specify C level
314coverage, and have C<Configure> and the F<Makefile> do all the right things
315automatically.
316
02f21748 317=head2 Make Config.pm cope with differences between built and installed perl
0bdfc961 318
319Quite often vendors ship a perl binary compiled with their (pay-for)
320compilers. People install a free compiler, such as gcc. To work out how to
321build extensions, Perl interrogates C<%Config>, so in this situation
322C<%Config> describes compilers that aren't there, and extension building
323fails. This forces people into choosing between re-compiling perl themselves
324using the compiler they have, or only using modules that the vendor ships.
325
326It would be good to find a way teach C<Config.pm> about the installation setup,
327possibly involving probing at install time or later, so that the C<%Config> in
328a binary distribution better describes the installed machine, when the
329installed machine differs from the build machine in some significant way.
330
728f4ecd 331=head2 linker specification files
332
333Some platforms mandate that you provide a list of a shared library's external
334symbols to the linker, so the core already has the infrastructure in place to
335do this for generating shared perl libraries. My understanding is that the
336GNU toolchain can accept an optional linker specification file, and restrict
337visibility just to symbols declared in that file. It would be good to extend
338F<makedef.pl> to support this format, and to provide a means within
339C<Configure> to enable it. This would allow Unix users to test that the
340export list is correct, and to build a perl that does not pollute the global
341namespace with private symbols.
342
a229ae3b 343=head2 Cross-compile support
344
345Currently C<Configure> understands C<-Dusecrosscompile> option. This option
346arranges for building C<miniperl> for TARGET machine, so this C<miniperl> is
347assumed then to be copied to TARGET machine and used as a replacement of full
348C<perl> executable.
349
d1307786 350This could be done little differently. Namely C<miniperl> should be built for
a229ae3b 351HOST and then full C<perl> with extensions should be compiled for TARGET.
d1307786 352This, however, might require extra trickery for %Config: we have one config
87a942b1 353first for HOST and then another for TARGET. Tools like MakeMaker will be
354mightily confused. Having around two different types of executables and
355libraries (HOST and TARGET) makes life interesting for Makefiles and
356shell (and Perl) scripts. There is $Config{run}, normally empty, which
357can be used as an execution wrapper. Also note that in some
358cross-compilation/execution environments the HOST and the TARGET do
359not see the same filesystem(s), the $Config{run} may need to do some
360file/directory copying back and forth.
0bdfc961 361
8537f021 362=head2 roffitall
363
364Make F<pod/roffitall> be updated by F<pod/buildtoc>.
365
98fca0e8 366=head2 Split "linker" from "compiler"
367
368Right now, Configure probes for two commands, and sets two variables:
369
370=over 4
371
b91dd380 372=item * C<cc> (in F<cc.U>)
98fca0e8 373
374This variable holds the name of a command to execute a C compiler which
375can resolve multiple global references that happen to have the same
376name. Usual values are F<cc> and F<gcc>.
377Fervent ANSI compilers may be called F<c89>. AIX has F<xlc>.
378
b91dd380 379=item * C<ld> (in F<dlsrc.U>)
98fca0e8 380
381This variable indicates the program to be used to link
382libraries for dynamic loading. On some systems, it is F<ld>.
383On ELF systems, it should be C<$cc>. Mostly, we'll try to respect
384the hint file setting.
385
386=back
387
8d159ec1 388There is an implicit historical assumption from around Perl5.000alpha
389something, that C<$cc> is also the correct command for linking object files
390together to make an executable. This may be true on Unix, but it's not true
391on other platforms, and there are a maze of work arounds in other places (such
392as F<Makefile.SH>) to cope with this.
98fca0e8 393
394Ideally, we should create a new variable to hold the name of the executable
395linker program, probe for it in F<Configure>, and centralise all the special
396case logic there or in hints files.
397
398A small bikeshed issue remains - what to call it, given that C<$ld> is already
8d159ec1 399taken (arguably for the wrong thing now, but on SunOS 4.1 it is the command
400for creating dynamically-loadable modules) and C<$link> could be confused with
401the Unix command line executable of the same name, which does something
402completely different. Andy Dougherty makes the counter argument "In parrot, I
403tried to call the command used to link object files and libraries into an
404executable F<link>, since that's what my vaguely-remembered DOS and VMS
405experience suggested. I don't think any real confusion has ensued, so it's
406probably a reasonable name for perl5 to use."
98fca0e8 407
408"Alas, I've always worried that introducing it would make things worse,
409since now the module building utilities would have to look for
410C<$Config{link}> and institute a fall-back plan if it weren't found."
8d159ec1 411Although I can see that as confusing, given that C<$Config{d_link}> is true
412when (hard) links are available.
98fca0e8 413
75585ce3 414=head2 Configure Windows using PowerShell
415
416Currently, Windows uses hard-coded config files based to build the
417config.h for compiling Perl. Makefiles are also hard-coded and need to be
418hand edited prior to building Perl. While this makes it easy to create a perl.exe
419that works across multiple Windows versions, being able to accurately
420configure a perl.exe for a specific Windows versions and VS C++ would be
421a nice enhancement. With PowerShell available on Windows XP and up, this
422may now be possible. Step 1 might be to investigate whether this is possible
423and use this to clean up our current makefile situation. Step 2 would be to
424see if there would be a way to use our existing metaconfig units to configure a
425Windows Perl or whether we go in a separate direction and make it so. Of
426course, we all know what step 3 is.
427
ab45a0fa 428=head2 decouple -g and -DDEBUGGING
429
430Currently F<Configure> automatically adds C<-DDEBUGGING> to the C compiler
431flags if it spots C<-g> in the optimiser flags. The pre-processor directive
eeab323f 432C<DEBUGGING> enables F<perl>'s command line C<-D> options, but in the process
ab45a0fa 433makes F<perl> slower. It would be good to disentangle this logic, so that
434C-level debugging with C<-g> and Perl level debugging with C<-D> can easily
435be enabled independently.
436
0bdfc961 437=head1 Tasks that need a little C knowledge
438
439These tasks would need a little C knowledge, but don't need any specific
440background or experience with XS, or how the Perl interpreter works
441
3d826b29 442=head2 Weed out needless PERL_UNUSED_ARG
443
444The C code uses the macro C<PERL_UNUSED_ARG> to stop compilers warning about
445unused arguments. Often the arguments can't be removed, as there is an
446external constraint that determines the prototype of the function, so this
447approach is valid. However, there are some cases where C<PERL_UNUSED_ARG>
448could be removed. Specifically
449
450=over 4
451
452=item *
453
454The prototypes of (nearly all) static functions can be changed
455
456=item *
457
458Unused arguments generated by short cut macros are wasteful - the short cut
459macro used can be changed.
460
461=back
462
fbf638cb 463=head2 Modernize the order of directories in @INC
464
465The way @INC is laid out by default, one cannot upgrade core (dual-life)
466modules without overwriting files. This causes problems for binary
3d14fd97 467package builders. One possible proposal is laid out in this
468message:
469L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2002-04/msg02380.html>.
fbf638cb 470
bcbaa2d5 471=head2 -Duse32bit*
472
473Natively 64-bit systems need neither -Duse64bitint nor -Duse64bitall.
474On these systems, it might be the default compilation mode, and there
475is currently no guarantee that passing no use64bitall option to the
476Configure process will build a 32bit perl. Implementing -Duse32bit*
477options would be nice for perl 5.12.
478
fee0a0f7 479=head2 Profile Perl - am I hot or not?
62403a3c 480
fee0a0f7 481The Perl source code is stable enough that it makes sense to profile it,
482identify and optimise the hotspots. It would be good to measure the
483performance of the Perl interpreter using free tools such as cachegrind,
484gprof, and dtrace, and work to reduce the bottlenecks they reveal.
485
486As part of this, the idea of F<pp_hot.c> is that it contains the I<hot> ops,
487the ops that are most commonly used. The idea is that by grouping them, their
488object code will be adjacent in the executable, so they have a greater chance
489of already being in the CPU cache (or swapped in) due to being near another op
490already in use.
62403a3c 491
492Except that it's not clear if these really are the most commonly used ops. So
fee0a0f7 493as part of exercising your skills with coverage and profiling tools you might
494want to determine what ops I<really> are the most commonly used. And in turn
495suggest evictions and promotions to achieve a better F<pp_hot.c>.
62403a3c 496
91d0cbf6 497One piece of Perl code that might make a good testbed is F<installman>.
498
98fed0ad 499=head2 Allocate OPs from arenas
500
501Currently all new OP structures are individually malloc()ed and free()d.
502All C<malloc> implementations have space overheads, and are now as fast as
503custom allocates so it would both use less memory and less CPU to allocate
504the various OP structures from arenas. The SV arena code can probably be
505re-used for this.
506
539f2c54 507Note that Configuring perl with C<-Accflags=-DPL_OP_SLAB_ALLOC> will use
508Perl_Slab_alloc() to pack optrees into a contiguous block, which is
509probably superior to the use of OP arenas, esp. from a cache locality
510standpoint. See L<Profile Perl - am I hot or not?>.
511
a229ae3b 512=head2 Improve win32/wince.c
0bdfc961 513
a229ae3b 514Currently, numerous functions look virtually, if not completely,
02f21748 515identical in both C<win32/wince.c> and C<win32/win32.c> files, which can't
6d71adcd 516be good.
517
c5b31784 518=head2 Use secure CRT functions when building with VC8 on Win32
519
520Visual C++ 2005 (VC++ 8.x) deprecated a number of CRT functions on the basis
521that they were "unsafe" and introduced differently named secure versions of
522them as replacements, e.g. instead of writing
523
524 FILE* f = fopen(__FILE__, "r");
525
526one should now write
527
528 FILE* f;
529 errno_t err = fopen_s(&f, __FILE__, "r");
530
531Currently, the warnings about these deprecations have been disabled by adding
532-D_CRT_SECURE_NO_DEPRECATE to the CFLAGS. It would be nice to remove that
533warning suppressant and actually make use of the new secure CRT functions.
534
535There is also a similar issue with POSIX CRT function names like fileno having
536been deprecated in favour of ISO C++ conformant names like _fileno. These
26a6faa8 537warnings are also currently suppressed by adding -D_CRT_NONSTDC_NO_DEPRECATE. It
c5b31784 538might be nice to do as Microsoft suggest here too, although, unlike the secure
539functions issue, there is presumably little or no benefit in this case.
540
038ae9a4 541=head2 Fix POSIX::access() and chdir() on Win32
542
543These functions currently take no account of DACLs and therefore do not behave
544correctly in situations where access is restricted by DACLs (as opposed to the
545read-only attribute).
546
547Furthermore, POSIX::access() behaves differently for directories having the
548read-only attribute set depending on what CRT library is being used. For
549example, the _access() function in the VC6 and VC7 CRTs (wrongly) claim that
550such directories are not writable, whereas in fact all directories are writable
551unless access is denied by DACLs. (In the case of directories, the read-only
552attribute actually only means that the directory cannot be deleted.) This CRT
553bug is fixed in the VC8 and VC9 CRTs (but, of course, the directory may still
554not actually be writable if access is indeed denied by DACLs).
555
556For the chdir() issue, see ActiveState bug #74552:
557http://bugs.activestate.com/show_bug.cgi?id=74552
558
559Therefore, DACLs should be checked both for consistency across CRTs and for
560the correct answer.
561
562(Note that perl's -w operator should not be modified to check DACLs. It has
563been written so that it reflects the state of the read-only attribute, even
564for directories (whatever CRT is being used), for symmetry with chmod().)
565
16815324 566=head2 strcat(), strcpy(), strncat(), strncpy(), sprintf(), vsprintf()
567
568Maybe create a utility that checks after each libperl.a creation that
569none of the above (nor sprintf(), vsprintf(), or *SHUDDER* gets())
570ever creep back to libperl.a.
571
572 nm libperl.a | ./miniperl -alne '$o = $F[0] if /:$/; print "$o $F[1]" if $F[0] eq "U" && $F[1] =~ /^(?:strn?c(?:at|py)|v?sprintf|gets)$/'
573
574Note, of course, that this will only tell whether B<your> platform
575is using those naughty interfaces.
576
de96509d 577=head2 -D_FORTIFY_SOURCE=2, -fstack-protector
578
579Recent glibcs support C<-D_FORTIFY_SOURCE=2> and recent gcc
580(4.1 onwards?) supports C<-fstack-protector>, both of which give
581protection against various kinds of buffer overflow problems.
582These should probably be used for compiling Perl whenever available,
583Configure and/or hints files should be adjusted to probe for the
584availability of these features and enable them as appropriate.
16815324 585
8964cfe0 586=head2 Arenas for GPs? For MAGIC?
587
588C<struct gp> and C<struct magic> are both currently allocated by C<malloc>.
589It might be a speed or memory saving to change to using arenas. Or it might
590not. It would need some suitable benchmarking first. In particular, C<GP>s
591can probably be changed with minimal compatibility impact (probably nothing
592outside of the core, or even outside of F<gv.c> allocates them), but they
593probably aren't allocated/deallocated often enough for a speed saving. Whereas
594C<MAGIC> is allocated/deallocated more often, but in turn, is also something
595more externally visible, so changing the rules here may bite external code.
596
3880c8ec 597=head2 Shared arenas
598
599Several SV body structs are now the same size, notably PVMG and PVGV, PVAV and
600PVHV, and PVCV and PVFM. It should be possible to allocate and return same
601sized bodies from the same actual arena, rather than maintaining one arena for
602each. This could save 4-6K per thread, of memory no longer tied up in the
603not-yet-allocated part of an arena.
604
8964cfe0 605
6d71adcd 606=head1 Tasks that need a knowledge of XS
607
608These tasks would need C knowledge, and roughly the level of knowledge of
609the perl API that comes from writing modules that use XS to interface to
610C.
611
e851c105 612=head2 Write an XS cookbook
613
614Create pod/perlxscookbook.pod with short, task-focused 'recipes' in XS that
615demonstrate common tasks and good practices. (Some of these might be
616extracted from perlguts.) The target audience should be XS novices, who need
617more examples than perlguts but something less overwhelming than perlapi.
618Recipes should provide "one pretty good way to do it" instead of TIMTOWTDI.
619
5b7d14ff 620Rather than focusing on interfacing Perl to C libraries, such a cookbook
621should probably focus on how to optimize Perl routines by re-writing them
622in XS. This will likely be more motivating to those who mostly work in
623Perl but are looking to take the next step into XS.
624
625Deconstructing and explaining some simpler XS modules could be one way to
626bootstrap a cookbook. (List::Util? Class::XSAccessor? Tree::Ternary_XS?)
627Another option could be deconstructing the implementation of some simpler
628functions in op.c.
629
05fb4e20 630=head2 Allow XSUBs to inline themselves as OPs
631
632For a simple XSUB, often the subroutine dispatch takes more time than the
633XSUB itself. The tokeniser already has the ability to inline constant
634subroutines - it would be good to provide a way to inline other subroutines.
635
636Specifically, simplest approach looks to be to allow an XSUB to provide an
637alternative implementation of itself as a custom OP. A new flag bit in
638C<CvFLAGS()> would signal to the peephole optimiser to take an optree
639such as this:
640
641 b <@> leave[1 ref] vKP/REFC ->(end)
642 1 <0> enter ->2
643 2 <;> nextstate(main 1 -e:1) v:{ ->3
644 a <2> sassign vKS/2 ->b
645 8 <1> entersub[t2] sKS/TARG,1 ->9
646 - <1> ex-list sK ->8
647 3 <0> pushmark s ->4
648 4 <$> const(IV 1) sM ->5
649 6 <1> rv2av[t1] lKM/1 ->7
650 5 <$> gv(*a) s ->6
651 - <1> ex-rv2cv sK ->-
652 7 <$> gv(*x) s/EARLYCV ->8
653 - <1> ex-rv2sv sKRM*/1 ->a
654 9 <$> gvsv(*b) s ->a
655
656perform the symbol table lookup of C<rv2cv> and C<gv(*x)>, locate the
657pointer to the custom OP that provides the direct implementation, and re-
658write the optree something like:
659
660 b <@> leave[1 ref] vKP/REFC ->(end)
661 1 <0> enter ->2
662 2 <;> nextstate(main 1 -e:1) v:{ ->3
663 a <2> sassign vKS/2 ->b
664 7 <1> custom_x -> 8
665 - <1> ex-list sK ->7
666 3 <0> pushmark s ->4
667 4 <$> const(IV 1) sM ->5
668 6 <1> rv2av[t1] lKM/1 ->7
669 5 <$> gv(*a) s ->6
670 - <1> ex-rv2cv sK ->-
671 - <$> ex-gv(*x) s/EARLYCV ->7
672 - <1> ex-rv2sv sKRM*/1 ->a
673 8 <$> gvsv(*b) s ->a
674
675I<i.e.> the C<gv(*)> OP has been nulled and spliced out of the execution
676path, and the C<entersub> OP has been replaced by the custom op.
677
678This approach should provide a measurable speed up to simple XSUBs inside
679tight loops. Initially one would have to write the OP alternative
680implementation by hand, but it's likely that this should be reasonably
681straightforward for the type of XSUB that would benefit the most. Longer
682term, once the run-time implementation is proven, it should be possible to
683progressively update ExtUtils::ParseXS to generate OP implementations for
684some XSUBs.
685
318bf708 686=head2 Remove the use of SVs as temporaries in dump.c
687
688F<dump.c> contains debugging routines to dump out the contains of perl data
689structures, such as C<SV>s, C<AV>s and C<HV>s. Currently, the dumping code
690B<uses> C<SV>s for its temporary buffers, which was a logical initial
691implementation choice, as they provide ready made memory handling.
692
693However, they also lead to a lot of confusion when it happens that what you're
694trying to debug is seen by the code in F<dump.c>, correctly or incorrectly, as
695a temporary scalar it can use for a temporary buffer. It's also not possible
696to dump scalars before the interpreter is properly set up, such as during
697ithreads cloning. It would be good to progressively replace the use of scalars
698as string accumulation buffers with something much simpler, directly allocated
699by C<malloc>. The F<dump.c> code is (or should be) only producing 7 bit
700US-ASCII, so output character sets are not an issue.
701
702Producing and proving an internal simple buffer allocation would make it easier
703to re-write the internals of the PerlIO subsystem to avoid using C<SV>s for
704B<its> buffers, use of which can cause problems similar to those of F<dump.c>,
705at similar times.
706
5d96f598 707=head2 safely supporting POSIX SA_SIGINFO
708
709Some years ago Jarkko supplied patches to provide support for the POSIX
710SA_SIGINFO feature in Perl, passing the extra data to the Perl signal handler.
711
712Unfortunately, it only works with "unsafe" signals, because under safe
713signals, by the time Perl gets to run the signal handler, the extra
714information has been lost. Moreover, it's not easy to store it somewhere,
715as you can't call mutexs, or do anything else fancy, from inside a signal
716handler.
717
718So it strikes me that we could provide safe SA_SIGINFO support
719
720=over 4
721
722=item 1
723
724Provide global variables for two file descriptors
725
726=item 2
727
728When the first request is made via C<sigaction> for C<SA_SIGINFO>, create a
729pipe, store the reader in one, the writer in the other
730
731=item 3
732
733In the "safe" signal handler (C<Perl_csighandler()>/C<S_raise_signal()>), if
734the C<siginfo_t> pointer non-C<NULL>, and the writer file handle is open,
735
736=over 8
737
738=item 1
739
740serialise signal number, C<struct siginfo_t> (or at least the parts we care
741about) into a small auto char buff
742
743=item 2
744
745C<write()> that (non-blocking) to the writer fd
746
747=over 12
748
749=item 1
750
751if it writes 100%, flag the signal in a counter of "signals on the pipe" akin
752to the current per-signal-number counts
753
754=item 2
755
756if it writes 0%, assume the pipe is full. Flag the data as lost?
757
758=item 3
759
760if it writes partially, croak a panic, as your OS is broken.
761
762=back
763
764=back
765
766=item 4
767
768in the regular C<PERL_ASYNC_CHECK()> processing, if there are "signals on
769the pipe", read the data out, deserialise, build the Perl structures on
770the stack (code in C<Perl_sighandler()>, the "unsafe" handler), and call as
771usual.
772
773=back
774
775I think that this gets us decent C<SA_SIGINFO> support, without the current risk
776of running Perl code inside the signal handler context. (With all the dangers
777of things like C<malloc> corruption that that currently offers us)
778
779For more information see the thread starting with this message:
780http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-03/msg00305.html
781
6d71adcd 782=head2 autovivification
783
784Make all autovivification consistent w.r.t LVALUE/RVALUE and strict/no strict;
785
786This task is incremental - even a little bit of work on it will help.
787
788=head2 Unicode in Filenames
789
790chdir, chmod, chown, chroot, exec, glob, link, lstat, mkdir, open,
791opendir, qx, readdir, readlink, rename, rmdir, stat, symlink, sysopen,
792system, truncate, unlink, utime, -X. All these could potentially accept
793Unicode filenames either as input or output (and in the case of system
794and qx Unicode in general, as input or output to/from the shell).
795Whether a filesystem - an operating system pair understands Unicode in
796filenames varies.
797
798Known combinations that have some level of understanding include
799Microsoft NTFS, Apple HFS+ (In Mac OS 9 and X) and Apple UFS (in Mac
800OS X), NFS v4 is rumored to be Unicode, and of course Plan 9. How to
801create Unicode filenames, what forms of Unicode are accepted and used
802(UCS-2, UTF-16, UTF-8), what (if any) is the normalization form used,
803and so on, varies. Finding the right level of interfacing to Perl
804requires some thought. Remember that an OS does not implicate a
805filesystem.
806
807(The Windows -C command flag "wide API support" has been at least
808temporarily retired in 5.8.1, and the -C has been repurposed, see
809L<perlrun>.)
810
87a942b1 811Most probably the right way to do this would be this:
812L</"Virtualize operating system access">.
813
6d71adcd 814=head2 Unicode in %ENV
815
816Currently the %ENV entries are always byte strings.
87a942b1 817See L</"Virtualize operating system access">.
6d71adcd 818
1f2e7916 819=head2 Unicode and glob()
820
821Currently glob patterns and filenames returned from File::Glob::glob()
87a942b1 822are always byte strings. See L</"Virtualize operating system access">.
1f2e7916 823
6d71adcd 824=head2 use less 'memory'
825
826Investigate trade offs to switch out perl's choices on memory usage.
827Particularly perl should be able to give memory back.
828
829This task is incremental - even a little bit of work on it will help.
830
831=head2 Re-implement C<:unique> in a way that is actually thread-safe
832
833The old implementation made bad assumptions on several levels. A good 90%
834solution might be just to make C<:unique> work to share the string buffer
835of SvPVs. That way large constant strings can be shared between ithreads,
836such as the configuration information in F<Config>.
837
838=head2 Make tainting consistent
839
840Tainting would be easier to use if it didn't take documented shortcuts and
841allow taint to "leak" everywhere within an expression.
842
843=head2 readpipe(LIST)
844
845system() accepts a LIST syntax (and a PROGRAM LIST syntax) to avoid
846running a shell. readpipe() (the function behind qx//) could be similarly
847extended.
848
6d71adcd 849=head2 Audit the code for destruction ordering assumptions
850
851Change 25773 notes
852
853 /* Need to check SvMAGICAL, as during global destruction it may be that
854 AvARYLEN(av) has been freed before av, and hence the SvANY() pointer
855 is now part of the linked list of SV heads, rather than pointing to
856 the original body. */
857 /* FIXME - audit the code for other bugs like this one. */
858
859adding the C<SvMAGICAL> check to
860
861 if (AvARYLEN(av) && SvMAGICAL(AvARYLEN(av))) {
862 MAGIC *mg = mg_find (AvARYLEN(av), PERL_MAGIC_arylen);
863
864Go through the core and look for similar assumptions that SVs have particular
865types, as all bets are off during global destruction.
866
749904bf 867=head2 Extend PerlIO and PerlIO::Scalar
868
869PerlIO::Scalar doesn't know how to truncate(). Implementing this
870would require extending the PerlIO vtable.
871
872Similarly the PerlIO vtable doesn't know about formats (write()), or
873about stat(), or chmod()/chown(), utime(), or flock().
874
875(For PerlIO::Scalar it's hard to see what e.g. mode bits or ownership
876would mean.)
877
878PerlIO doesn't do directories or symlinks, either: mkdir(), rmdir(),
879opendir(), closedir(), seekdir(), rewinddir(), glob(); symlink(),
880readlink().
881
94da6c29 882See also L</"Virtualize operating system access">.
883
3236f110 884=head2 -C on the #! line
885
886It should be possible to make -C work correctly if found on the #! line,
887given that all perl command line options are strict ASCII, and -C changes
888only the interpretation of non-ASCII characters, and not for the script file
889handle. To make it work needs some investigation of the ordering of function
890calls during startup, and (by implication) a bit of tweaking of that order.
891
d6c1e11f 892=head2 Organize error messages
893
894Perl's diagnostics (error messages, see L<perldiag>) could use
a8d0aeb9 895reorganizing and formalizing so that each error message has its
d6c1e11f 896stable-for-all-eternity unique id, categorized by severity, type, and
897subsystem. (The error messages would be listed in a datafile outside
c4bd451b 898of the Perl source code, and the source code would only refer to the
899messages by the id.) This clean-up and regularizing should apply
d6c1e11f 900for all croak() messages.
901
902This would enable all sorts of things: easier translation/localization
903of the messages (though please do keep in mind the caveats of
904L<Locale::Maketext> about too straightforward approaches to
905translation), filtering by severity, and instead of grepping for a
906particular error message one could look for a stable error id. (Of
907course, changing the error messages by default would break all the
908existing software depending on some particular error message...)
909
910This kind of functionality is known as I<message catalogs>. Look for
911inspiration for example in the catgets() system, possibly even use it
912if available-- but B<only> if available, all platforms will B<not>
de96509d 913have catgets().
d6c1e11f 914
915For the really pure at heart, consider extending this item to cover
916also the warning messages (see L<perllexwarn>, C<warnings.pl>).
3236f110 917
0bdfc961 918=head1 Tasks that need a knowledge of the interpreter
3298bd4d 919
0bdfc961 920These tasks would need C knowledge, and knowledge of how the interpreter works,
921or a willingness to learn.
3298bd4d 922
de6375e3 923=head2 truncate() prototype
924
925The prototype of truncate() is currently C<$$>. It should probably
926be C<*$> instead. (This is changed in F<opcode.pl>)
927
2d0587d8 928=head2 decapsulation of smart match argument
929
930Currently C<$foo ~~ $object> will die with the message "Smart matching a
931non-overloaded object breaks encapsulation". It would be nice to allow
932to bypass this by using explictly the syntax C<$foo ~~ %$object> or
933C<$foo ~~ @$object>.
934
565590b5 935=head2 error reporting of [$a ; $b]
936
937Using C<;> inside brackets is a syntax error, and we don't propose to change
938that by giving it any meaning. However, it's not reported very helpfully:
939
940 $ perl -e '$a = [$b; $c];'
941 syntax error at -e line 1, near "$b;"
942 syntax error at -e line 1, near "$c]"
943 Execution of -e aborted due to compilation errors.
944
945It should be possible to hook into the tokeniser or the lexer, so that when a
946C<;> is parsed where it is not legal as a statement terminator (ie inside
947C<{}> used as a hashref, C<[]> or C<()>) it issues an error something like
948I<';' isn't legal inside an expression - if you need multiple statements use a
949do {...} block>. See the thread starting at
950http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2008-09/msg00573.html
951
718140ec 952=head2 lexicals used only once
953
954This warns:
955
956 $ perl -we '$pie = 42'
957 Name "main::pie" used only once: possible typo at -e line 1.
958
959This does not:
960
961 $ perl -we 'my $pie = 42'
962
963Logically all lexicals used only once should warn, if the user asks for
d6f4ea2e 964warnings. An unworked RT ticket (#5087) has been open for almost seven
965years for this discrepancy.
718140ec 966
a3d15f9a 967=head2 UTF-8 revamp
968
969The handling of Unicode is unclean in many places. For example, the regexp
970engine matches in Unicode semantics whenever the string or the pattern is
971flagged as UTF-8, but that should not be dependent on an internal storage
e1b711da 972detail of the string.
a3d15f9a 973
974=head2 Properly Unicode safe tokeniser and pads.
975
976The tokeniser isn't actually very UTF-8 clean. C<use utf8;> is a hack -
977variable names are stored in stashes as raw bytes, without the utf-8 flag
978set. The pad API only takes a C<char *> pointer, so that's all bytes too. The
979tokeniser ignores the UTF-8-ness of C<PL_rsfp>, or any SVs returned from
980source filters. All this could be fixed.
981
636e63cb 982=head2 state variable initialization in list context
983
984Currently this is illegal:
985
986 state ($a, $b) = foo();
987
a2874905 988In Perl 6, C<state ($a) = foo();> and C<(state $a) = foo();> have different
a8d0aeb9 989semantics, which is tricky to implement in Perl 5 as currently they produce
a2874905 990the same opcode trees. The Perl 6 design is firm, so it would be good to
a8d0aeb9 991implement the necessary code in Perl 5. There are comments in
a2874905 992C<Perl_newASSIGNOP()> that show the code paths taken by various assignment
993constructions involving state variables.
636e63cb 994
4fedb12c 995=head2 Implement $value ~~ 0 .. $range
996
997It would be nice to extend the syntax of the C<~~> operator to also
998understand numeric (and maybe alphanumeric) ranges.
a393eb28 999
1000=head2 A does() built-in
1001
1002Like ref(), only useful. It would call the C<DOES> method on objects; it
1003would also tell whether something can be dereferenced as an
1004array/hash/etc., or used as a regexp, etc.
1005L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-03/msg00481.html>
1006
1007=head2 Tied filehandles and write() don't mix
1008
1009There is no method on tied filehandles to allow them to be called back by
1010formats.
4fedb12c 1011
53967bb9 1012=head2 Propagate compilation hints to the debugger
1013
1014Currently a debugger started with -dE on the command-line doesn't see the
1015features enabled by -E. More generally hints (C<$^H> and C<%^H>) aren't
1016propagated to the debugger. Probably it would be a good thing to propagate
1017hints from the innermost non-C<DB::> scope: this would make code eval'ed
1018in the debugger see the features (and strictures, etc.) currently in
1019scope.
1020
d10fc472 1021=head2 Attach/detach debugger from running program
1626a787 1022
cd793d32 1023The old perltodo notes "With C<gdb>, you can attach the debugger to a running
1024program if you pass the process ID. It would be good to do this with the Perl
0bdfc961 1025debugger on a running Perl program, although I'm not sure how it would be
1026done." ssh and screen do this with named pipes in /tmp. Maybe we can too.
1626a787 1027
0bdfc961 1028=head2 LVALUE functions for lists
1029
1030The old perltodo notes that lvalue functions don't work for list or hash
1031slices. This would be good to fix.
1032
0bdfc961 1033=head2 regexp optimiser optional
1034
1035The regexp optimiser is not optional. It should configurable to be, to allow
1036its performance to be measured, and its bugs to be easily demonstrated.
1037
02f21748 1038=head2 delete &function
1039
1040Allow to delete functions. One can already undef them, but they're still
1041in the stash.
1042
ef36c6a7 1043=head2 C</w> regex modifier
1044
1045That flag would enable to match whole words, and also to interpolate
1046arrays as alternations. With it, C</P/w> would be roughly equivalent to:
1047
1048 do { local $"='|'; /\b(?:P)\b/ }
1049
1050See L<http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-01/msg00400.html>
1051for the discussion.
1052
0bdfc961 1053=head2 optional optimizer
1054
1055Make the peephole optimizer optional. Currently it performs two tasks as
1056it walks the optree - genuine peephole optimisations, and necessary fixups of
1057ops. It would be good to find an efficient way to switch out the
1058optimisations whilst keeping the fixups.
1059
1060=head2 You WANT *how* many
1061
1062Currently contexts are void, scalar and list. split has a special mechanism in
1063place to pass in the number of return values wanted. It would be useful to
1064have a general mechanism for this, backwards compatible and little speed hit.
1065This would allow proposals such as short circuiting sort to be implemented
1066as a module on CPAN.
1067
1068=head2 lexical aliases
1069
1070Allow lexical aliases (maybe via the syntax C<my \$alias = \$foo>.
1071
1072=head2 entersub XS vs Perl
1073
1074At the moment pp_entersub is huge, and has code to deal with entering both
1075perl and XS subroutines. Subroutine implementations rarely change between
1076perl and XS at run time, so investigate using 2 ops to enter subs (one for
1077XS, one for perl) and swap between if a sub is redefined.
2810d901 1078
de535794 1079=head2 Self-ties
2810d901 1080
de535794 1081Self-ties are currently illegal because they caused too many segfaults. Maybe
a8d0aeb9 1082the causes of these could be tracked down and self-ties on all types
de535794 1083reinstated.
0bdfc961 1084
1085=head2 Optimize away @_
1086
1087The old perltodo notes "Look at the "reification" code in C<av.c>".
1088
87a942b1 1089=head2 Virtualize operating system access
1090
1091Implement a set of "vtables" that virtualizes operating system access
1092(open(), mkdir(), unlink(), readdir(), getenv(), etc.) At the very
1093least these interfaces should take SVs as "name" arguments instead of
1094bare char pointers; probably the most flexible and extensible way
e1a3d5d1 1095would be for the Perl-facing interfaces to accept HVs. The system
1096needs to be per-operating-system and per-file-system
1097hookable/filterable, preferably both from XS and Perl level
87a942b1 1098(L<perlport/"Files and Filesystems"> is good reading at this point,
1099in fact, all of L<perlport> is.)
1100
e1a3d5d1 1101This has actually already been implemented (but only for Win32),
1102take a look at F<iperlsys.h> and F<win32/perlhost.h>. While all Win32
1103variants go through a set of "vtables" for operating system access,
e1020413 1104non-Win32 systems currently go straight for the POSIX/Unix-style
e1a3d5d1 1105system/library call. Similar system as for Win32 should be
1106implemented for all platforms. The existing Win32 implementation
1107probably does not need to survive alongside this proposed new
1108implementation, the approaches could be merged.
87a942b1 1109
1110What would this give us? One often-asked-for feature this would
94da6c29 1111enable is using Unicode for filenames, and other "names" like %ENV,
1112usernames, hostnames, and so forth.
1113(See L<perlunicode/"When Unicode Does Not Happen">.)
1114
1115But this kind of virtualization would also allow for things like
1116virtual filesystems, virtual networks, and "sandboxes" (though as long
1117as dynamic loading of random object code is allowed, not very safe
1118sandboxes since external code of course know not of Perl's vtables).
1119An example of a smaller "sandbox" is that this feature can be used to
1120implement per-thread working directories: Win32 already does this.
1121
1122See also L</"Extend PerlIO and PerlIO::Scalar">.
87a942b1 1123
ac6197af 1124=head2 Investigate PADTMP hash pessimisation
1125
9a2f2e6b 1126The peephole optimiser converts constants used for hash key lookups to shared
057163d7 1127hash key scalars. Under ithreads, something is undoing this work.
ac6197af 1128See http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2007-09/msg00793.html
1129
057163d7 1130=head2 Store the current pad in the OP slab allocator
1131
1132=for clarification
1133I hope that I got that "current pad" part correct
1134
1135Currently we leak ops in various cases of parse failure. I suggested that we
1136could solve this by always using the op slab allocator, and walking it to
1137free ops. Dave comments that as some ops are already freed during optree
1138creation one would have to mark which ops are freed, and not double free them
1139when walking the slab. He notes that one problem with this is that for some ops
1140you have to know which pad was current at the time of allocation, which does
1141change. I suggested storing a pointer to the current pad in the memory allocated
1142for the slab, and swapping to a new slab each time the pad changes. Dave thinks
1143that this would work.
1144
52960e22 1145=head2 repack the optree
1146
1147Repacking the optree after execution order is determined could allow
057163d7 1148removal of NULL ops, and optimal ordering of OPs with respect to cache-line
1149filling. The slab allocator could be reused for this purpose. I think that
1150the best way to do this is to make it an optional step just before the
1151completed optree is attached to anything else, and to use the slab allocator
1152unchanged, so that freeing ops is identical whether or not this step runs.
1153Note that the slab allocator allocates ops downwards in memory, so one would
1154have to actually "allocate" the ops in reverse-execution order to get them
1155contiguous in memory in execution order.
1156
1157See http://www.nntp.perl.org/group/perl.perl5.porters/2007/12/msg131975.html
1158
1159Note that running this copy, and then freeing all the old location ops would
1160cause their slabs to be freed, which would eliminate possible memory wastage if
1161the previous suggestion is implemented, and we swap slabs more frequently.
52960e22 1162
12e06b6f 1163=head2 eliminate incorrect line numbers in warnings
1164
1165This code
1166
1167 use warnings;
1168 my $undef;
1169
1170 if ($undef == 3) {
1171 } elsif ($undef == 0) {
1172 }
1173
18a16cc5 1174used to produce this output:
12e06b6f 1175
1176 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1177 Use of uninitialized value in numeric eq (==) at wrong.pl line 4.
1178
18a16cc5 1179where the line of the second warning was misreported - it should be line 5.
1180Rafael fixed this - the problem arose because there was no nextstate OP
1181between the execution of the C<if> and the C<elsif>, hence C<PL_curcop> still
1182reports that the currently executing line is line 4. The solution was to inject
1183a nextstate OPs for each C<elsif>, although it turned out that the nextstate
1184OP needed to be a nulled OP, rather than a live nextstate OP, else other line
1185numbers became misreported. (Jenga!)
12e06b6f 1186
1187The problem is more general than C<elsif> (although the C<elsif> case is the
1188most common and the most confusing). Ideally this code
1189
1190 use warnings;
1191 my $undef;
1192
1193 my $a = $undef + 1;
1194 my $b
1195 = $undef
1196 + 1;
1197
1198would produce this output
1199
1200 Use of uninitialized value $undef in addition (+) at wrong.pl line 4.
1201 Use of uninitialized value $undef in addition (+) at wrong.pl line 7.
1202
1203(rather than lines 4 and 5), but this would seem to require every OP to carry
1204(at least) line number information.
1205
1206What might work is to have an optional line number in memory just before the
1207BASEOP structure, with a flag bit in the op to say whether it's present.
1208Initially during compile every OP would carry its line number. Then add a late
1209pass to the optimiser (potentially combined with L</repack the optree>) which
1210looks at the two ops on every edge of the graph of the execution path. If
1211the line number changes, flags the destination OP with this information.
1212Once all paths are traced, replace every op with the flag with a
1213nextstate-light op (that just updates C<PL_curcop>), which in turn then passes
1214control on to the true op. All ops would then be replaced by variants that
1215do not store the line number. (Which, logically, why it would work best in
1216conjunction with L</repack the optree>, as that is already copying/reallocating
1217all the OPs)
1218
18a16cc5 1219(Although I should note that we're not certain that doing this for the general
1220case is worth it)
1221
52960e22 1222=head2 optimize tail-calls
1223
1224Tail-calls present an opportunity for broadly applicable optimization;
1225anywhere that C<< return foo(...) >> is called, the outer return can
1226be replaced by a goto, and foo will return directly to the outer
1227caller, saving (conservatively) 25% of perl's call&return cost, which
1228is relatively higher than in C. The scheme language is known to do
1229this heavily. B::Concise provides good insight into where this
1230optimization is possible, ie anywhere entersub,leavesub op-sequence
1231occurs.
1232
1233 perl -MO=Concise,-exec,a,b,-main -e 'sub a{ 1 }; sub b {a()}; b(2)'
1234
1235Bottom line on this is probably a new pp_tailcall function which
1236combines the code in pp_entersub, pp_leavesub. This should probably
1237be done 1st in XS, and using B::Generate to patch the new OP into the
1238optrees.
1239
0bdfc961 1240=head1 Big projects
1241
1242Tasks that will get your name mentioned in the description of the "Highlights
87a942b1 1243of 5.12"
0bdfc961 1244
1245=head2 make ithreads more robust
1246
4e577f8b 1247Generally make ithreads more robust. See also L</iCOW>
0bdfc961 1248
1249This task is incremental - even a little bit of work on it will help, and
1250will be greatly appreciated.
1251
6c047da7 1252One bit would be to write the missing code in sv.c:Perl_dirp_dup.
1253
59c7f7d5 1254Fix Perl_sv_dup, et al so that threads can return objects.
1255
0bdfc961 1256=head2 iCOW
1257
1258Sarathy and Arthur have a proposal for an improved Copy On Write which
1259specifically will be able to COW new ithreads. If this can be implemented
1260it would be a good thing.
1261
1262=head2 (?{...}) closures in regexps
1263
1264Fix (or rewrite) the implementation of the C</(?{...})/> closures.
1265
1266=head2 A re-entrant regexp engine
1267
1268This will allow the use of a regex from inside (?{ }), (??{ }) and
1269(?(?{ })|) constructs.
6bda09f9 1270
6bda09f9 1271=head2 Add class set operations to regexp engine
1272
1273Apparently these are quite useful. Anyway, Jeffery Friedl wants them.
1274
1275demerphq has this on his todo list, but right at the bottom.
44a7a252 1276
1277
1278=head1 Tasks for microperl
1279
1280
1281[ Each and every one of these may be obsolete, but they were listed
1282 in the old Todo.micro file]
1283
1284
1285=head2 make creating uconfig.sh automatic
1286
1287=head2 make creating Makefile.micro automatic
1288
1289=head2 do away with fork/exec/wait?
1290
1291(system, popen should be enough?)
1292
1293=head2 some of the uconfig.sh really needs to be probed (using cc) in buildtime:
1294
1295(uConfigure? :-) native datatype widths and endianness come to mind
1296