Father Chrysostomos [Sun, 15 Sep 2013 02:16:22 +0000 (19:16 -0700)]
perldelta for #119161
Father Chrysostomos [Sun, 15 Sep 2013 01:55:13 +0000 (18:55 -0700)]
perldelta for
bf1b738b6/#118931
Father Chrysostomos [Sun, 15 Sep 2013 01:54:55 +0000 (18:54 -0700)]
Father Chrysostomos [Sun, 15 Sep 2013 01:49:03 +0000 (18:49 -0700)]
Father Chrysostomos [Sun, 15 Sep 2013 01:46:29 +0000 (18:46 -0700)]
perldelta for
bbfdc870/#118691
Father Chrysostomos [Sun, 15 Sep 2013 01:34:22 +0000 (18:34 -0700)]
perldelta for
ce0d59fdd1c/#7503,#118753
Father Chrysostomos [Sun, 15 Sep 2013 01:29:55 +0000 (18:29 -0700)]
perldelta for
1b268608/#118753
Father Chrysostomos [Sun, 15 Sep 2013 00:36:56 +0000 (17:36 -0700)]
perldelta for
4f62cd62/#118693
Father Chrysostomos [Sun, 15 Sep 2013 00:34:56 +0000 (17:34 -0700)]
perldelta for 3a6ce63
Father Chrysostomos [Sun, 15 Sep 2013 00:34:39 +0000 (17:34 -0700)]
Father Chrysostomos [Sun, 15 Sep 2013 00:29:37 +0000 (17:29 -0700)]
perldelta: missing fullstop
Father Chrysostomos [Sun, 15 Sep 2013 00:27:50 +0000 (17:27 -0700)]
perl5200delta: Devel::NYTProf is fixed
5.06 works with bleadperl.
Father Chrysostomos [Sun, 15 Sep 2013 00:23:11 +0000 (17:23 -0700)]
Reduce false positives for @hsh{$s} and @ary[$s] warnings
This resolves tickets #28380 and #114024.
Commit
95a31aad5 did something similar to this for the new %hash{...}
syntax. This commit extends it to @ slices and combines the two
code paths.
The heuristics in toke.c can easily produce false positives. So the
op is flagged as being a candidate for the warning. Then when op.c
has the op tree available, it examines it to see whether the heuristic
may have been a false positive.
This avoids bugs with qw "foo bar baz" and sub calls triggering
the warning.
The source code is no longer available for the warning, so we recon-
struct it from the op tree, skipping the subscript if it is anything
other than a const op.
This means that @hash{$foo} comes out as @hash{...} and @hash{foo} as
@hash{"foo"}. It also meeans that @hash{"]"} is displayed correctly
instead of as @hash{"].
Commit
95a31aad5 also modified the heuristic for %hash{...} to exempt
qw altogether. But it did not exempt it if it was preceded by a tab.
So this commit rectifies that.
This commit also improves the false positive detection by exempting
any ops returning lists that can get past toke.c’s heuristic. I went
through the entire list of ops, but I may have missed some.
Also, @ slices on the lhs of = are exempt, as they change the context
and are hence actually useful.
Father Chrysostomos [Sat, 14 Sep 2013 18:14:42 +0000 (11:14 -0700)]
dump.c: Dump SLICEWARNING flag for kv[ah]slice, too
I accidentally omitted this in commit
95a31aad5.
Father Chrysostomos [Sat, 14 Sep 2013 01:25:02 +0000 (18:25 -0700)]
Move ‘Scalar value better written as’ tests from toke to op
I am about to move the warning itself from toke.c to op.c. Moving the
tests, too, makes them easier to find for future maintainters who see
it in op.c.
Father Chrysostomos [Fri, 13 Sep 2013 20:49:33 +0000 (13:49 -0700)]
op.c: Factor out common varname code
Steve Hay [Sat, 14 Sep 2013 12:46:21 +0000 (13:46 +0100)]
Upgrade Encode from version 2.54 to 2.55
Steve Hay [Fri, 13 Sep 2013 22:19:09 +0000 (23:19 +0100)]
Remove a __BORLANDC__ stowaway
Support for __BORLANDC__ was axed (actually, chainsawed) in
378eeda70c.
Steve Hay [Fri, 13 Sep 2013 21:57:59 +0000 (22:57 +0100)]
Fix the VC6 build on Windows following commit
3be8f09452
The VC6 compiler has slightly flaky support for __int64 and chokes on util.c
when using the 64-bit implementation of drand48, so switch it to the 32-bit
version instead.
Craig A. Berry [Fri, 13 Sep 2013 19:59:01 +0000 (14:59 -0500)]
fix O_CREAT without O_TRUNC in cpan/autodie/t/utf8_open.t
Awaiting upstream application as [CPAN RT#87237].
Father Chrysostomos [Fri, 13 Sep 2013 20:13:33 +0000 (13:13 -0700)]
Test kv slices in places that expect %{}
Father Chrysostomos [Fri, 13 Sep 2013 09:10:21 +0000 (02:10 -0700)]
op.c:leave_scope: use mg_free before sv_force_normal
This is part of ticket #119295.
Commit 7fa949d (5.19.3) allowed copy-on-write with compile-time con-
stants. That caused this to fail:
use Variable::OnDestruct:
{
my $h = "foo";
on_destruct $h, sub { warn defined $_[0] ? $_[0] : "undef" };
$x++; # prev statement must not be last in the block
}
It prints undef instead of foo.
It turns out this is much older:
use Variable::OnDestruct;
{
my $h = __PACKAGE__;
on_destruct $h, sub { warn defined $_[0] ? $_[0] : "undef" };
$x++; # prev statement must not be last in the block
}
This one prints undef starting with 5.17.3 (
a6d7a4ac1).
But even before that, this prints undef:
use Variable::OnDestruct;
{
my $h = \1;
on_destruct $h, sub { warn defined $_[0] ? $_[0] : "undef" };
$x++; # prev statement must not be last in the block
}
In all these cases, the scalar is getting undefined before free magic
triggers (Variable::OnDestruct uses free magic).
Usually when a scalar is freed, the magic is triggered before anything
else. When a lexical scalar is ‘freed’ on scope exit (cleared for
reuse on scope entry), the order is different. References, globs and
copy-on-write scalars become undefined (via sv_force_normal) before
magic is triggered.
There is no reason for the order to be different here, and it causes
unpredictable behaviour (you never know when you will or will not have
a cow). So change the order in scope exit to match regular freeing.
Nicholas Clark [Tue, 10 Sep 2013 08:32:24 +0000 (10:32 +0200)]
#ifdef DEBUGGING, not QDEBUGGING, in regcomp.c
Father Chrysostomos [Fri, 13 Sep 2013 08:25:55 +0000 (01:25 -0700)]
[Merge] Key/value hash and array slices
Father Chrysostomos [Fri, 13 Sep 2013 01:13:42 +0000 (18:13 -0700)]
Document the %hash{$scalar} warning
Father Chrysostomos [Thu, 12 Sep 2013 21:30:41 +0000 (14:30 -0700)]
Fewer false positives for %hash{$scalar} warning
Instead of warning in the lexer, flag the op and then warn in op.c,
when the op tree is available, so we don’t end up warning for actual
lists or for sub calls.
Also, only warn in scalar context, as in list context $hash{$scalar}
and %hash{$scalar} do different things.
In op.c we no longer have easy access to the source code, so recon-
struct the hash/array access based on the op tree. This means
%hash{foo} becomes %hash{"foo"}. We only reconstruct constant keys,
so %hash{++$x} becomes %hash{...}. This also corrects erroneous
dumps, like %hash{"} for %hash{"}"}.
Instead of triggering the warning solely based on the op tree, we
still keep the heuristic in toke.c, so that common workarounds for
that warning (e.g., {q<key>} and {("key")}) continue to work.
The heuristic in toke.c is tweaked to avoid warning for qw().
In a future commit I plan to extend this to the existing @array[0] and
@hash{key} warnings, to avoid false positives.
Father Chrysostomos [Sun, 8 Sep 2013 23:11:29 +0000 (16:11 -0700)]
pp.c: Allow 64-bit stack offsets in kv[ah]slice
Father Chrysostomos [Sun, 8 Sep 2013 22:50:46 +0000 (15:50 -0700)]
Improve new slice docs in perldata.pod
Father Chrysostomos [Sun, 8 Sep 2013 21:55:34 +0000 (14:55 -0700)]
perldata.pod: Capitalise =head3 consistently
Father Chrysostomos [Sun, 8 Sep 2013 21:54:40 +0000 (14:54 -0700)]
perldata.pod: consistent spaces after dots
Father Chrysostomos [Sun, 8 Sep 2013 21:51:27 +0000 (14:51 -0700)]
pp.c:pp_kvhslice: remove unused var
Father Chrysostomos [Sun, 8 Sep 2013 21:50:00 +0000 (14:50 -0700)]
Increase $B::Deparse::VERSION to 1.23
Father Chrysostomos [Sun, 8 Sep 2013 21:49:24 +0000 (14:49 -0700)]
Increase $Opcode::VERSION to 1.26
Father Chrysostomos [Sun, 8 Sep 2013 21:48:26 +0000 (14:48 -0700)]
Add new opcodes to Opcode.pm
Father Chrysostomos [Sun, 8 Sep 2013 21:38:24 +0000 (14:38 -0700)]
Sort MANIFEST
Father Chrysostomos [Sun, 8 Sep 2013 21:19:34 +0000 (14:19 -0700)]
Use diag_listed_as two suppress two diag.t failures
Father Chrysostomos [Sun, 8 Sep 2013 21:19:13 +0000 (14:19 -0700)]
Add two missing perldiag entries
Father Chrysostomos [Sun, 8 Sep 2013 21:09:23 +0000 (14:09 -0700)]
Hard-code op name in ck_exists|delete errors
The reason is manifold:
• ck_exists and ck_delete are each only ever called with one op type,
so we can hard-code the op, which takes less code.
• The messages in perldiag are more helpful if they include
the op name.
• Changing them in op.c too avoids the need for
diag_listed_as clutter.
(I know, two of these are not listed in perldiag yet. I’m get-
ting to that.)
Father Chrysostomos [Sun, 8 Sep 2013 20:53:23 +0000 (13:53 -0700)]
Suppress a warning in kvhslice.t
The bugginess of the ‘Scalar value better written as...’ warning
applies to %new{silce} syntax, too.
Ruslan Zakirov [Sun, 8 Sep 2013 18:13:06 +0000 (22:13 +0400)]
update manifest
Ruslan Zakirov [Sun, 8 Sep 2013 15:40:34 +0000 (19:40 +0400)]
short documentation for key/value hash/array slices
Ruslan Zakirov [Sat, 7 Sep 2013 17:41:45 +0000 (21:41 +0400)]
test index/value array slices
Ruslan Zakirov [Tue, 12 Mar 2013 13:00:48 +0000 (17:00 +0400)]
test key/value hash slices
Ruslan Zakirov [Sat, 2 Mar 2013 12:57:41 +0000 (16:57 +0400)]
index/value array slice operation
kvaslice operator that imlements %a[0,2,4] syntax which
result in list of index/value pairs. Implemented in
consistency with "key/value hash slice" operator.
Ruslan Zakirov [Sun, 24 Mar 2013 11:53:12 +0000 (15:53 +0400)]
scalar value %x{'a'} better written as $x{'a'}
Ruslan Zakirov [Wed, 27 Feb 2013 15:39:39 +0000 (19:39 +0400)]
key/value hash slice operation
kvhslice operator that implements %h{1,2,3,4} syntax which
returns list of key value pairs rather than just values
(regular slices).
Nicholas Clark [Wed, 11 Sep 2013 09:26:29 +0000 (11:26 +0200)]
Move BmUSEFUL to the IV slot from the NV slot, simplifying union _xnvu.
This change is made possible by commit
8922e4382e9c1488 (June 2013), which
eliminated BmPREVIOUS, which had been using the IV slot.
Father Chrysostomos [Fri, 13 Sep 2013 01:26:30 +0000 (18:26 -0700)]
Preserve undef identity in const ops
This fixes the one remaining issue in ticket #105906.
Under ithreads, the constants are transferred to the pad at compile
time. Since the pads are AVs, and since AVs used &PL_sv_undef to rep-
resent unused slots, &PL_sv_undef could not be stored in the pad, and
so was copied.
That meant the same constant would produce a different scalar at each
call site:
$ ./perl -Ilib -le 'BEGIN { $::{z} = \undef } for(z,z) { print \$_ }'
SCALAR(0x7fecb38062e0)
SCALAR(0x7fecb3806100)
Commit ce0d59f changed the way AVs work, so now we *can* store
&PL_sv_undef in an AV meaningfully, and there is no longer any need
for this copying hack.
Ricardo Signes [Fri, 13 Sep 2013 01:36:54 +0000 (21:36 -0400)]
convert mktables to parent.pm instead of base.pm
Tony Cook [Fri, 13 Sep 2013 01:51:45 +0000 (11:51 +1000)]
Tony Cook [Fri, 13 Sep 2013 01:34:16 +0000 (11:34 +1000)]
[perl #115928] use a consistent internal rand on all platforms
Tony Cook [Tue, 10 Sep 2013 00:09:43 +0000 (10:09 +1000)]
[perl #115928] we don't use drand48_r or random_r any longer
Removing this should mean that metaconfig will remove the units from
the built Configure
Tony Cook [Mon, 9 Sep 2013 04:06:35 +0000 (14:06 +1000)]
[perl #115928] Configure now selects our internal drand48()
Tony Cook [Thu, 29 Aug 2013 04:52:15 +0000 (14:52 +1000)]
revert config_h.SH changes, now handled by Configure
Tony Cook [Mon, 9 Sep 2013 04:44:57 +0000 (14:44 +1000)]
[perl #115928] a consistent (public) rand() implementation
Based on Yves's random branch work.
This version makes the new random number visible to external modules,
for example, List::Util's XS shuffle() implementation.
I've also added a 64-bit implementation when HAS_QUAD is true, this
should be significantly faster, even on 32-bit CPUs. This is intended to
produce exactly the same sequence as the original implementation.
The original version of this commit retained the "freebsd" name from
Yves's original work for the function and data structure names. I've
removed "freebsd" from most function names so the name isn't an issue
if we choose to replace the implementation,
Ricardo Signes [Mon, 9 Sep 2013 02:26:10 +0000 (22:26 -0400)]
perlreapi: use parent in example, not base
Ricardo Signes [Mon, 9 Sep 2013 02:07:03 +0000 (22:07 -0400)]
overload tests: use parent instead of base
Ricardo Signes [Mon, 9 Sep 2013 02:06:57 +0000 (22:06 -0400)]
XS-Typemap tests: use parent instead of base
Ricardo Signes [Mon, 9 Sep 2013 02:06:21 +0000 (22:06 -0400)]
B: use parent instead of base
Ricardo Signes [Mon, 9 Sep 2013 02:01:47 +0000 (22:01 -0400)]
Pod::Html: use parent instead of base
Ricardo Signes [Mon, 9 Sep 2013 02:01:19 +0000 (22:01 -0400)]
Hash::Util::FieldHash: use parent instead of base
Ricardo Signes [Mon, 9 Sep 2013 02:00:46 +0000 (22:00 -0400)]
FileCache: use parent instead of base
Matthew Horsfall [Thu, 12 Sep 2013 18:26:45 +0000 (14:26 -0400)]
Allow semi-colon on TYPEMAP: line to work without it on the closing delim.
This means:
TYPEMAP: <<FOO;
Something T_IV
FOO
will now work.
Steve Hay [Thu, 12 Sep 2013 21:10:56 +0000 (22:10 +0100)]
perldelta - IO and File::Find have been upgraded
Steve Hay [Thu, 12 Sep 2013 20:57:09 +0000 (21:57 +0100)]
Upgrade perlfaq from version 5.0150043 to 5.0150044
Slaven Rezic [Thu, 12 Sep 2013 12:04:27 +0000 (14:04 +0200)]
File::Find: diagnostics in case of a failed chdir
Father Chrysostomos [Thu, 12 Sep 2013 00:53:03 +0000 (17:53 -0700)]
perldiag: Rewrap an entry for better splain output
John P. Linderman [Thu, 12 Sep 2013 14:15:32 +0000 (16:15 +0200)]
Correct the citation for Peter McIlroy's sorting paper.
Also update John P. Linderman's e-mail address.
Nicholas Clark [Thu, 12 Sep 2013 14:22:51 +0000 (16:22 +0200)]
Update John P. Linderman's e-mail address in AUTHORS.
Steve Hay [Thu, 12 Sep 2013 07:53:43 +0000 (08:53 +0100)]
Upgrade autodie from version 2.20 to 2.21
One change pushed upstream has been assimilated, the other not yet.
Steve Hay [Wed, 11 Sep 2013 19:52:44 +0000 (20:52 +0100)]
SADAHIRO confirms that Unicode::(Collate|Normalize) can be UPSTREAM=>'cpan'
Steve Hay [Wed, 11 Sep 2013 19:51:04 +0000 (20:51 +0100)]
Upgrade Module::Metadata from version 1.000017 to 1.000018
The taint.t test can now be included since it no longer uses Test::Fatal.
Steve Hay [Wed, 11 Sep 2013 09:29:09 +0000 (10:29 +0100)]
EXCLUDE Module-Metadata's t/taint.t: we don't have Test::Fatal in core
Steve Hay [Wed, 11 Sep 2013 08:39:17 +0000 (09:39 +0100)]
Upgrade Perl-OSType from version 1.004 to 1.005
Steve Hay [Wed, 11 Sep 2013 08:34:33 +0000 (09:34 +0100)]
Upgrade Module::Metadata from version 1.000016 to 1.000017
Tony Cook [Wed, 11 Sep 2013 13:13:56 +0000 (23:13 +1000)]
[perl #88814] sigdispatch.t work on openbsd 5.2
Karl Williamson [Wed, 11 Sep 2013 03:28:31 +0000 (21:28 -0600)]
perlapi: Add doc for my_strlcpy, my_strlcat
Karl Williamson [Wed, 11 Sep 2013 02:40:25 +0000 (20:40 -0600)]
regcomp.c: Use Perl_form() instead of buggy sprintf
Commit
4cabb89a737018190d4e09360a6615e19160709f introduced tests
which include trying to look up a Unicode property whose name
is an upper Latin-range character. This caused errors and a segfault
on some but not all patforms. It also turns out that the error is in
some locales but not others.
The problem was there all along, and this commit merely exposed it; and
the bug is in the libc sprintf() for those platforms, which was used
to create the look-up property name. This version of sprintf() turns
out to be locale-sensitive, and in a UTF-8 locale, it refuses to format
a %s that isn't valid UTF-8. My guess is that sprintf and printf share
implementation, and such sensitivity may be warranted for printf, but
certainly not for sprintf. It is undocumented behavior. And it is
incorrect UTF-8 handling even if one were to output UTF-8 only. The
reason for that is any malformed text should be turned into the
REPLACEMENT CHARACTER, not just skipped over. This is a potential
security hole in in this sprintf() version.
The solution I ended up for Perl is to replace the sprintf() with
Perl_form(). I also looked at my_strlcpy() and friends. Neither one is
as convenient as the buggy sprintf. Feel free to replace this with
something better.
Karl Williamson [Mon, 25 Mar 2013 19:09:09 +0000 (13:09 -0600)]
Use separate macros for byte vs uv Unicode
This removes a macro not yet even in a development release, and splits
its calls into two classes: those where the input is a byte; and those
where it can be any unsigned integer. The byte implementation avoids a
function call on EBCDIC platforms.
Tony Cook [Wed, 11 Sep 2013 00:47:13 +0000 (10:47 +1000)]
[perl #118091] Split gv_fetchpvn_flags into smaller functions
Brian Fraser [Sun, 4 Aug 2013 17:55:56 +0000 (14:55 -0300)]
gv.c: Split part of find_default_stash into gv_is_in_main.
gv_is_in_main() checks if an unqualified identifier is in the main::
stash.
Brian Fraser [Sun, 4 Aug 2013 17:23:27 +0000 (14:23 -0300)]
gv.c: Rename magicalize_gv into gv_magicalize, make it more specific.
Namely, gv_magicalize no longer stores the GV into the stash, which
is gv_fetchpvn_flags' job.
Brian Fraser [Tue, 21 May 2013 14:32:45 +0000 (11:32 -0300)]
gv.c, gv_fetchpvn_flags: Split another chunk of magic-checking code.
This bit is called when a GV already exists, but it's name is length-one
and it's on the main:: stash, so it might have multiple kinds of magic,
like $! and %!, or @+ and %+.
Brian Fraser [Tue, 21 May 2013 14:25:09 +0000 (11:25 -0300)]
gv.c, gv_fetchpvn_flags: Comments for when the glob already exists.
Brian Fraser [Fri, 10 May 2013 07:17:58 +0000 (04:17 -0300)]
gv.c: Explain the purpose of the <none>:: stash a bit.
It's primarily a historical oddity. Its main purpose is to store
variables after an exception from 'use strict "vars"':
eval q{ use strict; $foo = 1 };
warn keys %{"<none>::"};
Brian Fraser [Thu, 9 May 2013 23:19:33 +0000 (20:19 -0300)]
gv_fetchpvn_flags: Simplify some warnings code
This line originally checked isLEXWARN_on and had two different code
paths depending on that -- one of which called ckWARN(), which already
checks for isLEXWARN_off, so rather than doing duplicate work, we
just call ckWARN().
Brian Fraser [Tue, 9 Apr 2013 08:28:18 +0000 (05:28 -0300)]
gv.c: Move the code that magicalizes new globs into magicalize_gv().
Brian Fraser [Tue, 9 Apr 2013 07:27:16 +0000 (04:27 -0300)]
gv.c: Begin splitting gv_fetchpvn_flags into smaller helper functions.
This commit takes a chunk of code out of gv_fetchpvn_flags and
turns it into two fuctions: parse_gv_stash_name and find_default_stash.
Ricardo Signes [Tue, 10 Sep 2013 22:32:10 +0000 (18:32 -0400)]
release schedule: placeholders for the future
David Golden [Tue, 10 Sep 2013 21:45:36 +0000 (17:45 -0400)]
perldiag: Fix severity of Unicode mapping warning
The "does not map to Unicode" warning severity was incorrect. It is
not fatal and can be controlled with the "utf8" category.
Steve Hay [Tue, 10 Sep 2013 21:26:31 +0000 (22:26 +0100)]
Oops, the HTTP::Tiny upgrade was 0.034->0.035, not 0.34->0.35
Steve Hay [Tue, 10 Sep 2013 21:17:55 +0000 (22:17 +0100)]
Upgrade HTTP::Tiny from version 0.34 o 0.35
EXCLUDE t/00-compile.t, which has been rewritten and no longer works in
core. (Other t/00-compile.t files from the same CPAN author are already
EXCLUDED.)
Brian Fraser [Tue, 10 Sep 2013 20:43:40 +0000 (21:43 +0100)]
Fix Windows compilation failure introduced by
946095af78
Patch taken from
Message-ID: <CA+nL+nad64D8hegnphMLnkLLR6dvcAj1X4eqV3CofMg=xdyu_Q@mail.gmail.com>
Smylers [Fri, 6 Sep 2013 10:56:07 +0000 (11:56 +0100)]
Added release manager test for copyright year
Smylers [Fri, 6 Sep 2013 10:07:27 +0000 (11:07 +0100)]
Test that README and perl -v copyright years match
Smylers [Fri, 6 Sep 2013 09:37:29 +0000 (10:37 +0100)]
Enhance runperl doc to mention its return value
Father Chrysostomos [Tue, 10 Sep 2013 15:36:43 +0000 (08:36 -0700)]
[Merge] Make regexp error messages utf8-clean
Brian Fraser [Thu, 29 Aug 2013 14:18:55 +0000 (11:18 -0300)]
Make the "Switch condition not recognized" error message UTF-8 clean
Brian Fraser [Fri, 30 Aug 2013 16:10:16 +0000 (13:10 -0300)]
t/re/reg_mesg.t: Tests for latin1 error messages/warnings
Brian Fraser [Fri, 6 Sep 2013 00:07:22 +0000 (21:07 -0300)]
regcomp.c: Introduce vFAIL2utf8f to replace the APPLY(X,Y) hack
The APPLY() hack was added in the previous commit to get
vFAIL4("%"UTF8f, UTF8fARG(a,b,c)) working. Without it,
vFAIL4 would complain about missing arguments, since it only
saw one, but if replaced with vFAIL2, then Simple_vFAIL2 would
complain about having too many arguments.
This commit introduces a vFAIL2utf8f macro that works around
this and enables us to remove the hack.