Robin Houston [Mon, 26 Dec 2005 17:56:48 +0000]
Test that constant overloading is propagated into eval
Message-ID: <
20051226175648.GA28402@rpc142.cs.man.ac.uk>
p4raw-id: //depot/perl@26521
Yitzchak Scott-Thoennes [Mon, 26 Dec 2005 18:42:40 +0000]
[DOC PATCH] rindex clarification
Message-ID: <
20051227024240.GA2940@efn.org>
plus another formatting nit
p4raw-id: //depot/perl@26520
Craig A. Berry [Tue, 27 Dec 2005 17:59:35 +0000]
fortify Pod::Simple::Search against non-case-preserving filesystems
From: "Craig A. Berry" <craigberry@mac.com>
Message-id: <
43B1D567.9080504@mac.com>
p4raw-id: //depot/perl@26519
Nicholas Clark [Thu, 29 Dec 2005 00:19:01 +0000]
A GVs stash can be NULL, so don't call macros that assume otherwise
without checking.
p4raw-id: //depot/perl@26518
Nicholas Clark [Wed, 28 Dec 2005 19:40:26 +0000]
As the backreferences AV doesn't hold references on its contents,
surely it should have AvREAL turned off?
p4raw-id: //depot/perl@26517
Nicholas Clark [Wed, 28 Dec 2005 13:35:03 +0000]
PL_amagic_generation is always non-zero, so remove the test for this.
(Since change 17990 added version object overloading)
p4raw-id: //depot/perl@26516
Steve Peters [Wed, 28 Dec 2005 13:31:02 +0000]
Upgrade to Sys-Syslog-0.11
p4raw-id: //depot/perl@26515
Gisle Aas [Wed, 28 Dec 2005 13:25:30 +0000]
Cleanup; replace a croak("panic: ...") with assert().
p4raw-id: //depot/perl@26514
Gisle Aas [Wed, 28 Dec 2005 13:12:30 +0000]
assert(av); Cleanup the way that the av_*() functions check
that they are really passed an av argument.
Should improve the test coverage a bit.
p4raw-id: //depot/perl@26513
Gisle Aas [Wed, 28 Dec 2005 12:38:03 +0000]
Use memchr() instead of ninstr() to locate the end-of-line.
The call to memchr() will faster.
p4raw-id: //depot/perl@26512
Gisle Aas [Wed, 28 Dec 2005 12:18:58 +0000]
Leaner ninstr().
The compiled function ended up 37% smaller on Linux/gcc.
I ought to be faster as well, but I did not try to measure that.
p4raw-id: //depot/perl@26511
Gisle Aas [Wed, 28 Dec 2005 11:01:46 +0000]
Remove test for NUL ending in r?instr().
This test is actually harmful because we will search
for any garbage found past the empty search string.
Ref change 26509.
p4raw-id: //depot/perl@26510
Gisle Aas [Wed, 28 Dec 2005 10:57:46 +0000]
Fix 2 off-by-one errors in the call to ninstr().
This code managed to still work because of a bug
in how ninstr() treats empty search strings.
p4raw-id: //depot/perl@26509
Steve Peters [Wed, 28 Dec 2005 03:27:13 +0000]
Upgrade to PathTools-3.15
p4raw-id: //depot/perl@26508
Nicholas Clark [Tue, 27 Dec 2005 23:39:42 +0000]
s/%/%%/ in the package name so that it can be used safely in the
sprintf format string (and hence not need runtime %s interpolation)
p4raw-id: //depot/perl@26507
Nicholas Clark [Tue, 27 Dec 2005 22:19:28 +0000]
Pass in the full arguments to ExtUtils::Constant::ProxySubs
Prototype all the missing constants, so that parsing doesn't depend on
the phase of the moon. (Well, the system headers)
Add an option for making missing constants generate errors at read time
(which can be during constant folding). This isn't the default.
p4raw-id: //depot/perl@26506
Andy Lester [Tue, 27 Dec 2005 14:39:39 +0000]
pre-likely cleanup
Message-ID: <
20051227203939.GC1781@petdance.com>
Includes a small fix to the changes in tryAMAGICbinW_var() in pp.h.
p4raw-id: //depot/perl@26505
Steve Peters [Tue, 27 Dec 2005 18:36:19 +0000]
It'd be nice if I hadn't broken corecpan.pl. Its fixed now, though.
p4raw-id: //depot/perl@26504
Nicholas Clark [Tue, 27 Dec 2005 17:33:13 +0000]
gv_fetch_flags in newATTRSUB can actually be const.
Comment the logic behind the terms in the ternary that chooses the
flags value.
p4raw-id: //depot/perl@26503
Nicholas Clark [Tue, 27 Dec 2005 17:24:11 +0000]
As what we're doing is generating subroutines, which are methods, we
need to invalidate cached methods. For constant.pm we need to create
&Internals::inc_sub_generation
p4raw-id: //depot/perl@26502
Yves Orton [Fri, 23 Dec 2005 20:06:52 +0000]
[patch] ext/b/t/lint.t fails on win32
Message-ID: <
9b18b3110512231106k7ad395bfp8a0c05cd0f87fc61@mail.gmail.com>
p4raw-id: //depot/perl@26501
Nicholas Clark [Tue, 27 Dec 2005 12:02:57 +0000]
Refactor the require failure message generation to use 1 less temporary
SV, and to build as much text as possible with newSVpvf
p4raw-id: //depot/perl@26500
Nicholas Clark [Tue, 27 Dec 2005 00:29:33 +0000]
Removing MAKE_JOBS_FIFO from %ENV causes FreeBSD make to forget about
any -j flags. (And their implied disabling of backwards compatibility,
which is the real cause of the make disttest failure).
p4raw-id: //depot/perl@26499
Nicholas Clark [Mon, 26 Dec 2005 23:52:44 +0000]
tabs, not spaces, in MANIFEST. No, I can't remember why this is.
Or whether it could be relaxed.
p4raw-id: //depot/perl@26498
Nicholas Clark [Mon, 26 Dec 2005 23:50:14 +0000]
Remove the hard wired test numbers in the generated test script for
simple tests. Use this script twice, firstly for testing the old style
autoloaded constants, then for testing the ProxySubs.
Make testing the dogfood/regeneration code optional, as the ProxySubs
output doesn't contain it.
p4raw-id: //depot/perl@26497
Nicholas Clark [Mon, 26 Dec 2005 23:48:00 +0000]
Support type UNDEF, and correct a bug in the macro generation by using
macro_from_item.
p4raw-id: //depot/perl@26496
Nicholas Clark [Mon, 26 Dec 2005 23:47:01 +0000]
Add method macro_from_item to encapsulate the entire logic for getting
the correct macro direct from the item hash.
p4raw-id: //depot/perl@26495
Nicholas Clark [Mon, 26 Dec 2005 23:10:53 +0000]
Don't try to store PL_sv_undef in pads. (For now)
p4raw-id: //depot/perl@26494
Steve Peters [Mon, 26 Dec 2005 22:01:49 +0000]
Upgrade to CPAN-1.80_56
p4raw-id: //depot/perl@26493
Steve Peters [Mon, 26 Dec 2005 17:29:13 +0000]
In this, the last tale of the NGROUPS saga, a former pumpking prods
a mere committer to remove the last of the NGROUPS-sized arrays...
Perl_magic_set() was using the last of these arrays to do the
lvalue work on $). Instead of an array, a pointer is used and
re-sized as needed.
p4raw-id: //depot/perl@26492
Nicholas Clark [Mon, 26 Dec 2005 17:01:35 +0000]
Add a couple of newlines for tidier regen output, consistent with the
code generated by WriteConstants()
p4raw-id: //depot/perl@26491
Nicholas Clark [Mon, 26 Dec 2005 16:13:57 +0000]
Add C_FH and XS_FH arguments to ExtUtils::Constant::WriteConstants,
to allow the caller to pass in file handles. Use this in Contant.t
with tied file handles to capture the output, rather than calling
the lower level routines directly.
p4raw-id: //depot/perl@26490
Jim Cromie [Mon, 26 Dec 2005 07:57:51 +0000]
chg 26486 removes last use of NGROUPS, so remove the macro-def
Message-ID: <
43B004EF.8060500@gmail.com>
p4raw-id: //depot/perl@26489
Nicholas Clark [Mon, 26 Dec 2005 12:48:10 +0000]
The values for proxy constant subroutines need to be read only.
p4raw-id: //depot/perl@26488
Nicholas Clark [Mon, 26 Dec 2005 12:31:01 +0000]
Rework constant.pm to take advantage of the space savings of proxy
constant subroutines whenever it can.
p4raw-id: //depot/perl@26487
Steve Peters [Mon, 26 Dec 2005 03:51:24 +0000]
Obviously, Perl_ingroup() is also using 256k of stack memory on Linux.
Adapt change #26480 to reduce memory usage here as well.
p4raw-link: @26480 on //depot/perl:
57d7c65eded7a5f963c5ce38ee196978a06e35df
p4raw-id: //depot/perl@26486
Nicholas Clark [Sat, 24 Dec 2005 16:27:28 +0000]
The value from caller doesn't change, so my $pkg = caller;
should be moved out of the loop.
p4raw-id: //depot/perl@26485
Nicholas Clark [Sat, 24 Dec 2005 16:06:10 +0000]
Tests for creating constants where prototypes or other symbols of the
same name already exist.
p4raw-id: //depot/perl@26484
Nicholas Clark [Sat, 24 Dec 2005 15:09:28 +0000]
Using a hash reference rather than a hash can save a hash copy.
p4raw-id: //depot/perl@26483
Nicholas Clark [Sat, 24 Dec 2005 14:48:42 +0000]
Unless the peephole optimiser already knows that we're in void context,
avoid even attempting the whole proxy constant sub copying
optimisation.
p4raw-id: //depot/perl@26482
Nicholas Clark [Sat, 24 Dec 2005 14:09:56 +0000]
Given that IO::Socket is documented as exporting all of Socket's
exports, there's no need for IO::Socket::INET or IO::Socket::UNIX to
use both. (And re-exporting removes some of the efficiency of the new
constants implementation)
p4raw-id: //depot/perl@26481
Jim Cromie [Fri, 23 Dec 2005 13:24:29 +0000]
Re: [perl #37907] Perl_magic_get reqires ridiculous amounts of stackmemory
Message-ID: <
43AC5CFD.3060200@gmail.com>
Includes a change in variable names from "j" to "num_groups".
p4raw-id: //depot/perl@26480
Nicholas Clark [Sat, 24 Dec 2005 00:01:50 +0000]
Looks like the macro CLK_TCK is not actually a constant on Solaris.
p4raw-id: //depot/perl@26479
Nicholas Clark [Fri, 23 Dec 2005 17:30:19 +0000]
For PERL_TRACK_MEMPOOL with PERL_POISON, Poison the end of any block
being shrunk in realloc()
p4raw-id: //depot/perl@26478
Nicholas Clark [Fri, 23 Dec 2005 17:24:44 +0000]
For the rare case of EMFILE during require, save object code space
by using newSVpvf
(Would it be worth using it for all require failures?)
p4raw-id: //depot/perl@26477
Nicholas Clark [Fri, 23 Dec 2005 16:55:35 +0000]
If PERL_TRACK_MEMPOOL and PERL_POISON are in use, then scribble all
over memory to invalidate it just before free()ing it.
p4raw-id: //depot/perl@26476
Nicholas Clark [Fri, 23 Dec 2005 16:20:53 +0000]
Provide support for types PVN and UNDEF in
ExtUtils::Constant::ProxySubs
p4raw-id: //depot/perl@26475
Nicholas Clark [Fri, 23 Dec 2005 15:27:10 +0000]
Create a struct to use as the header with PERL_TRACK_MEMPOOL, so that
other information can be stored in it.
p4raw-id: //depot/perl@26474
Nicholas Clark [Fri, 23 Dec 2005 13:54:37 +0000]
ExtUtils::ParseXS needs to stack $filepathname to make #line directives
in #INCLUDEs work.
p4raw-id: //depot/perl@26473
Nicholas Clark [Fri, 23 Dec 2005 12:21:19 +0000]
Add support for PV to ExtUtils::Constant::ProxySubs, and enable its
use in Sys::Syslog
p4raw-id: //depot/perl@26472
Nicholas Clark [Fri, 23 Dec 2005 12:01:29 +0000]
inlining autoloaded constants is TODOne.
p4raw-id: //depot/perl@26471
Nicholas Clark [Fri, 23 Dec 2005 11:36:43 +0000]
Now that proxy subroutines can be unproxied in 2 places without
becoming the same newCONSTSUB, need an explicit check to avoid a
warning about subroutines being redefined, as there has never been
a warning when you assing the same subroutine to a glob's GvCV()
p4raw-id: //depot/perl@26470
Nicholas Clark [Fri, 23 Dec 2005 11:35:21 +0000]
like and unlike weren't reporting failure where correctly in test.pl
p4raw-id: //depot/perl@26469
Rafael Garcia-Suarez [Fri, 23 Dec 2005 11:06:34 +0000]
Fix build of Compress::Zlib on MVS (by Paul Marquess)
p4raw-id: //depot/perl@26468
Rafael Garcia-Suarez [Fri, 23 Dec 2005 10:13:30 +0000]
Clarify example of .. in perlop
to fix [perl #38008] Documentation bug: perlop
plus some minor formatting nits
p4raw-id: //depot/perl@26467
chromatic [Thu, 22 Dec 2005 20:58:00 +0000]
[REPATCH lib/UNIVERSAL.pm] Recommend Against Using Methods as Functions
Message-Id: <
200512221158.54474.chromatic@wgz.org>
p4raw-id: //depot/perl@26466
Nicholas Clark [Fri, 23 Dec 2005 01:38:01 +0000]
Need to be more careful with the symbol table manipulation - if there
is already an entry in place, back off and generate a real constant
subroutine. (fixes lib/Net/hostent.t failure)
Fix a bug with how hv_store was being called.
p4raw-id: //depot/perl@26465
Nicholas Clark [Fri, 23 Dec 2005 01:09:05 +0000]
SIGRTMAX, SIGRTMIN, LDBL_MAX, LDBL_MIN, LDBL_EPSILON and HUGE_VAL
aren't constants on (at least some) Linux.
And more evil than that, HUGE_VAL likes to use the C preprocessor
in its value, necessitating use of a temporary.
p4raw-id: //depot/perl@26464
Nicholas Clark [Fri, 23 Dec 2005 00:37:04 +0000]
Use inlineable proxy constant subs for Socket
p4raw-id: //depot/perl@26463
Steve Peters [Fri, 23 Dec 2005 00:28:54 +0000]
ctime_r() and asctime_r() seem to have some problems before HP-UX 11.
p4raw-id: //depot/perl@26462
Andy Lester [Thu, 22 Dec 2005 16:00:44 +0000]
Speed up Perl_sv_derived_from
Message-ID: <
20051222220044.GH4370@petdance.com>
Date: Thu, 22 Dec 2005 16:00:44 -0600
p4raw-id: //depot/perl@26461
Nicholas Clark [Fri, 23 Dec 2005 00:18:49 +0000]
Create a sub LC_ALL with the correct prototype.
p4raw-id: //depot/perl@26460
Nicholas Clark [Fri, 23 Dec 2005 00:18:27 +0000]
De-duplicate the items in @EXPORT. (@EXPORT is built from all the tags,
and some constants are mentioned in multiple tags).
Should this logic be in Exporter::Heavy?
p4raw-id: //depot/perl@26459
Steve Peters [Fri, 23 Dec 2005 00:10:25 +0000]
Make hostname.t deal with undefined or blank hostnames and domain
names.
p4raw-id: //depot/perl@26458
Steve Peters [Fri, 23 Dec 2005 00:06:58 +0000]
Fix for pod2man() in ExtUtils::Command::MM to work with
the Pod::Man 2.04.
p4raw-id: //depot/perl@26457
Andy Lester [Thu, 22 Dec 2005 16:38:39 +0000]
Speedup (rn|n)?instr
Message-ID: <
20051222223839.GI4370@petdance.com>
Date: Thu, 22 Dec 2005 16:38:39 -0600
p4raw-id: //depot/perl@26456
Nicholas Clark [Thu, 22 Dec 2005 23:57:27 +0000]
Use inlineable proxy constant subs for POSIX.
There may be trouble ahead, as it seems that not all POSIX "constants"
are. I wonder if too many systems are going to have too many
variations to make this viable.
p4raw-id: //depot/perl@26455
Nicholas Clark [Thu, 22 Dec 2005 23:43:22 +0000]
Missing ; in Tie::File
p4raw-id: //depot/perl@26454
Nicholas Clark [Thu, 22 Dec 2005 21:02:19 +0000]
Make ExtUtils::Constant generate the inlineable proxy constant subs.
So far just enough to make Fcntl work.
p4raw-id: //depot/perl@26453
Rafael Garcia-Suarez [Thu, 22 Dec 2005 19:37:42 +0000]
Compilation help for EBCDIC platforms, from Jarkko,
tested by Rajarshi Das
p4raw-id: //depot/perl@26452
Nicholas Clark [Thu, 22 Dec 2005 19:00:52 +0000]
Break out the code to generate #ifdef/#endif into new methods
macro_to_ifdef and macro_to_endif
Add an args hashref to normalise_items; provide a
disable_utf8_duplication argument to disable the utf8 duplication code.
p4raw-id: //depot/perl@26451
Andy Lester [Thu, 22 Dec 2005 10:57:17 +0000]
Ho! Ho! Ho! Santa brings consting!
Message-ID: <
20051222165717.GA2874@petdance.com>
p4raw-id: //depot/perl@26450
Nicholas Clark [Thu, 22 Dec 2005 17:01:33 +0000]
Break out the item normalisation code into a method normalise_items.
p4raw-id: //depot/perl@26449
Nicholas Clark [Thu, 22 Dec 2005 15:43:20 +0000]
Regression tests for proxy subroutine glob assignment.
Fix a bug (it turns out that a typeglob isn't SvOK())
Remove stray debugging code.
p4raw-id: //depot/perl@26448
Robin Houston [Thu, 22 Dec 2005 11:54:55 +0000]
eliminate HINT_HH_FOR_EVAL
Message-ID: <
20051222115455.GB30301@rpc142.cs.man.ac.uk>
p4raw-id: //depot/perl@26447
Nicholas Clark [Thu, 22 Dec 2005 11:23:34 +0000]
Add an optimisation to allow proxy constant subroutines to be copied
as proxy constant subroutines in a new symbol table where possible.
(Rather than converting them to full blown constant subroutines and
instantiating 2 typeglobs)
p4raw-id: //depot/perl@26446
Rafael Garcia-Suarez [Thu, 22 Dec 2005 11:13:25 +0000]
Regenerate Changes
p4raw-id: //depot/perl@26445
Gisle Aas [Thu, 22 Dec 2005 10:42:36 +0000]
Remove unused variable 'stash' since change 26437.
p4raw-id: //depot/perl@26444
Rafael Garcia-Suarez [Thu, 22 Dec 2005 10:32:17 +0000]
Regenerate perlapi.pod.
Fix indentation of code examples in perlsyn.
p4raw-id: //depot/perl@26443
Robin Houston [Wed, 21 Dec 2005 13:51:18 +0000]
switch / ~~ doccos
Message-ID: <
20051221135118.GC26527@rpc142.cs.man.ac.uk>
p4raw-id: //depot/perl@26442
Jan Dubois [Wed, 21 Dec 2005 15:04:30 +0000]
wrong setting in canned win32/config.vc64 file
From: "Jan Dubois" <jand@ActiveState.com>
Message-ID: <
06c701c60682$e650adc0$d563a8c0@candy>
p4raw-id: //depot/perl@26441
Steve Peters [Thu, 22 Dec 2005 03:08:59 +0000]
Add documentation for method domainname() in Net::Domain. Add some
additional tests for Net::Domain. Also, since it is tested, remove
test case from lib/1_compile.t. One step closer to $500.
p4raw-id: //depot/perl@26440
Nicholas Clark [Wed, 21 Dec 2005 22:40:04 +0000]
Avoid negating an unsigned value. (The offset in the SV body table)
p4raw-id: //depot/perl@26439
Nicholas Clark [Wed, 21 Dec 2005 21:20:23 +0000]
For consistency sv_2cv should not leave the stash pointer uninitialised
for some cases. Although it seems that only pp_sort pays any attention
to it, of all the code on CPAN.
p4raw-id: //depot/perl@26438
Nicholas Clark [Wed, 21 Dec 2005 20:38:42 +0000]
Localise the stash variable in pp_entersub().
p4raw-id: //depot/perl@26437
Nicholas Clark [Wed, 21 Dec 2005 19:37:48 +0000]
Undo my goto spaghetti from change 18340 - what we really need are
hints to compilers about likey/unlikely branches.
p4raw-id: //depot/perl@26436
Steve Peters [Wed, 21 Dec 2005 19:04:37 +0000]
Add tests for untested math functions in POSIX
p4raw-id: //depot/perl@26435
Nicholas Clark [Wed, 21 Dec 2005 18:34:03 +0000]
The lref argument of sv_2cv is actually passed onwards to gv_fetchsv,
so it is a bitmap of flag bits rather than simple TRUE/FALSE.
p4raw-id: //depot/perl@26434
Rafael Garcia-Suarez [Wed, 21 Dec 2005 17:07:05 +0000]
Remove switch from perltodo
p4raw-id: //depot/perl@26433
Robin Houston [Wed, 21 Dec 2005 11:00:08 +0000]
Feature bundle is now :5.10, and add -E switch
Message-ID: <
20051221110008.GB25877@rpc142.cs.man.ac.uk>
p4raw-id: //depot/perl@26432
Rafael Garcia-Suarez [Wed, 21 Dec 2005 11:12:32 +0000]
Fix for [perl #37999] lc() + Latin-1 chars is failing erratically
based on copying part of change #22196 from do_chomp() to do_chop().
p4raw-link: @22196 on //depot/perl:
a6aa349da2cd706a05b205fa788c278b74c24bdc
p4raw-id: //depot/perl@26431
Jan Dubois [Tue, 20 Dec 2005 16:42:24 +0000]
Fix bug in change #26379
From: "Jan Dubois" <jand@ActiveState.com>
Message-ID: <
046001c605c7$69470290$d563a8c0@candy>
p4raw-id: //depot/perl@26430
Nicholas Clark [Tue, 20 Dec 2005 23:12:17 +0000]
rv_2av, rv_2cv and rv_2hv also make use of OPf_SPECIAL, so update the
descriptive comment to include them.
p4raw-id: //depot/perl@26429
Rafael Garcia-Suarez [Tue, 20 Dec 2005 20:50:26 +0000]
Remove some hardcoded references to ./perl in tests
p4raw-id: //depot/perl@26428
Nicholas Clark [Tue, 20 Dec 2005 20:13:12 +0000]
Take care in toke.c not to convert constant subroutine reference
proxies into full blown PVGVs with PVCVs, and recognise them and inline
their values. Adds a new function gv_const_sv(gv) to return the value
of the constant subroutine from a GV, and adds a cv parameter to
S_intuit_method.
p4raw-id: //depot/perl@26427
H.Merijn Brand [Tue, 20 Dec 2005 19:56:58 +0000]
Small update on using threads on HP-UX 10.20
64bit is far beyond experimental by now
p4raw-id: //depot/perl@26426
Nicholas Clark [Tue, 20 Dec 2005 19:55:46 +0000]
Add a GV_NOADD_NOINIT flag to gv_fetch{pv,pvn,sv} that disables
addition of new typeglobs, and also disables initialisation of any
typeglob placeholders. Needed to make the new constant subroutine
proxy references work efficiently.
p4raw-id: //depot/perl@26425
Rafael Garcia-Suarez [Tue, 20 Dec 2005 19:37:26 +0000]
Fix typos in regular expressions
p4raw-id: //depot/perl@26424
Nicholas Clark [Tue, 20 Dec 2005 16:34:27 +0000]
Get the "cv" from the "gv" once, and be more careful so that we can
cope if either aren't their regular types.
cv_const_sv should verify the type of SV passed, instead of assuming
that it's a viable CV.
p4raw-id: //depot/perl@26423
Nicholas Clark [Tue, 20 Dec 2005 15:11:09 +0000]
Croak if gv_init doesn't know how to create a typeglob from that type
of referant. Test that ARRAY, HASH, PVIO, CODE and FORMAT croak.
Globs are actually first class assignable objects, so test that you
can create a constant subroutine that returns one.
p4raw-id: //depot/perl@26422