This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Karl Williamson [Fri, 17 Oct 2014 03:07:11 +0000 (21:07 -0600)]
POSIX/t/export.t: Generalize for non-ASCII platforms
Karl Williamson [Fri, 17 Oct 2014 03:03:50 +0000 (21:03 -0600)]
Pod-Functions/Functions_pm.PL: Generalize for non-ASCII platforms
Karl Williamson [Fri, 17 Oct 2014 03:00:56 +0000 (21:00 -0600)]
Hash-Util tests: Generalize for non-ASCII platforms
Karl Williamson [Fri, 17 Oct 2014 02:50:29 +0000 (20:50 -0600)]
t/op/override.t: Generalize for non-ASCII platforms
Karl Williamson [Mon, 12 May 2014 01:34:08 +0000 (19:34 -0600)]
t/op/chars.t: Generalize to not use code page specific
In EBCDIC, \c? resolves to the APC character. We can use table lookup
to find out what that is on the current platform, without having to know
what code page we are using.
Karl Williamson [Thu, 27 Jun 2013 00:23:27 +0000 (18:23 -0600)]
regexp_unicode_prop.t: Generalize for non-ASCII platforms
Karl Williamson [Wed, 26 Jun 2013 21:25:37 +0000 (15:25 -0600)]
t/lib/feature/implicit: Generalize for EBCDIC platforms
Karl Williamson [Tue, 21 May 2013 04:01:02 +0000 (22:01 -0600)]
ext/XS-APItest/t/hash.t: Generalize to run on non-ASCII platforms
Karl Williamson [Tue, 21 May 2013 03:54:01 +0000 (21:54 -0600)]
t/lib/feature/implicit: Generalize for non-ASCII platforms
Karl Williamson [Fri, 19 Apr 2013 19:58:12 +0000 (13:58 -0600)]
t/op/coreamp.t: Generalize for non-ASCII platfomrs
Karl Williamson [Thu, 18 Apr 2013 03:49:10 +0000 (21:49 -0600)]
t/re/re_tests: Some tests are platform-specific
Karl Williamson [Wed, 17 Apr 2013 14:22:36 +0000 (08:22 -0600)]
t/io/crlf.t: Generalize for non-ASCII platforms
Karl Williamson [Tue, 16 Apr 2013 18:13:07 +0000 (12:13 -0600)]
ext/B/t/b.t: Generalize for non-ASCII platforms
Karl Williamson [Tue, 16 Apr 2013 17:50:04 +0000 (11:50 -0600)]
t/op/warn.t: Generalize for non-ASCII platforms
Karl Williamson [Tue, 16 Apr 2013 16:18:02 +0000 (10:18 -0600)]
re/reg_email.t: Generalize for non-ASCII platforms
This replaces all the hard-coded hex character values. It uses the new
(?[ ]) notation. I checked that the compiled regex matches the exact
same code points as before these changes.
Karl Williamson [Sat, 13 Apr 2013 22:19:20 +0000 (16:19 -0600)]
ext/XS-APItest/t/svpeek.t: Extend to EBCDIC platforms
Karl Williamson [Sat, 13 Apr 2013 22:14:35 +0000 (16:14 -0600)]
ext/XS-APItest/t/svpv_magic.t: Generalize for non-ASCII platforms
Karl Williamson [Sat, 13 Apr 2013 21:35:52 +0000 (15:35 -0600)]
lib/utf8.t: Generalize for non-ASCII platforms
This includes choosing a different code point that has 3 bytes in both
UTF-8 and UTF-EBCDIC, so that the pos numbers work for both.
Karl Williamson [Sat, 13 Apr 2013 21:16:44 +0000 (15:16 -0600)]
t/uni/parser.t: Extend to EBCDIC platforms
Karl Williamson [Sat, 13 Apr 2013 20:41:46 +0000 (14:41 -0600)]
t/uni/method.t: Extend to EBCDIC platforms
I couldn't figure out a way to not use the hard-coded values
Karl Williamson [Sat, 13 Apr 2013 20:26:09 +0000 (14:26 -0600)]
t/op/magic.t: Generalize for non-ASCII platforms
Karl Williamson [Sat, 13 Apr 2013 19:36:41 +0000 (13:36 -0600)]
t/io/through.t: Generalize for EBCDIC platforms
This uses hard-coded values for EBCDIC because of the shell issues
Karl Williamson [Sun, 7 Apr 2013 16:45:14 +0000 (10:45 -0600)]
t/op/goto.t: Generalize for non-ASCII platforms
Karl Williamson [Thu, 4 Apr 2013 02:15:17 +0000 (20:15 -0600)]
t/re/pat.t: Generalize for EBCDIC
Karl Williamson [Sat, 6 Apr 2013 05:34:50 +0000 (23:34 -0600)]
t/uni/overload.t: Generalize for non-ASCII platforms
Karl Williamson [Sat, 6 Apr 2013 05:34:13 +0000 (23:34 -0600)]
t/uni/method.t: Generalize for non-ASCII platforms
Karl Williamson [Sat, 6 Apr 2013 05:33:28 +0000 (23:33 -0600)]
t/op/utf8magic.t: Generalize for non-ASCII platforms
Karl Williamson [Sat, 6 Apr 2013 05:32:57 +0000 (23:32 -0600)]
t/op/evalbytes.t: Generalize for non-ASCII platforms
Karl Williamson [Sat, 6 Apr 2013 19:01:54 +0000 (13:01 -0600)]
t/op/utfhash.t: Fix to work on EBCDIC 1047
This .t thought it could have an __END__ in the middle of DATA input,
and the first read would read in up to that, and the second would read
the rest. In bisecting, I couldn't find a time when this ever worked.
Now this has a marker, and does a split on that marker, choosing the
first or second half depending.
Note that this only works on ASCII and EBCDIC 1047 platforms. It could
be extended for the other code pages Perl purportedly supports.
Karl Williamson [Tue, 2 Apr 2013 04:29:16 +0000 (22:29 -0600)]
t/re/pat_re_eval.t: Generalize for non-ASCII platforms
Karl Williamson [Tue, 2 Apr 2013 03:08:20 +0000 (21:08 -0600)]
t/op/split.t: EBCDIC fixes
Karl Williamson [Tue, 2 Apr 2013 02:43:03 +0000 (20:43 -0600)]
re/pat_advanced.t: EBCDIC fixes
This includes not skipping some EBCDIC that formerly was, since we now
have testing infrastructure that makes this easy.
Karl Williamson [Tue, 2 Apr 2013 02:01:04 +0000 (20:01 -0600)]
t/io/utf8.t: EBCDIC fixes
Karl Williamson [Sat, 30 Mar 2013 18:32:09 +0000 (12:32 -0600)]
t/uni/fold.t: Generalize for non-ASCII platforms
Karl Williamson [Wed, 27 Mar 2013 22:55:55 +0000 (16:55 -0600)]
lib/feature/bundle: Generalize for non-ASCII platforms
Karl Williamson [Wed, 27 Mar 2013 22:08:04 +0000 (16:08 -0600)]
XS-APItest/t/fetch_pad_names.t: Skip if EBCDIC
This could be ported, but there's a lot of stuff to convert
Karl Williamson [Wed, 27 Mar 2013 18:05:53 +0000 (12:05 -0600)]
ext/XS-APItest/t/utf8.t: Fix so passes EBCDIC
This involves skipping much of the tests, because it would be painful to
generalize for the different EBCDIC code pages.
Karl Williamson [Wed, 27 Mar 2013 17:27:06 +0000 (11:27 -0600)]
ext/re/t/re_funcs_u.t: Generalize for non-ASCII platforms
Karl Williamson [Wed, 27 Mar 2013 16:33:44 +0000 (10:33 -0600)]
t/op/print.t: Skip an EBCDIC test
This could be fixed (the values would probably change depending on the
code page), but the code that would get exercised is unlikely to vary
depending on character set.
Karl Williamson [Tue, 26 Mar 2013 02:43:38 +0000 (20:43 -0600)]
op/chop.t: Fix for EBCDIC
One test is skipped because the code point is not representable on
EBCDIC platforms. Other tests are modified to work on EBCDIC.
Karl Williamson [Tue, 26 Mar 2013 01:56:50 +0000 (19:56 -0600)]
t/op/lc.t: Generalize for non-ASCII platforms
This had code that attempted this, but it was wrong. The conversion to
EBCDIC must be done before the \U, or similar.
Karl Williamson [Mon, 25 Mar 2013 21:04:14 +0000 (15:04 -0600)]
Skip some porting tests under EBCDIC
EBCDIC won't work on these because of inherent differences from ASCII,
including a different sort order.
For porting/regen.t, we use the shipped parsing rules on EBCDIC platforms
Karl Williamson [Mon, 25 Mar 2013 20:59:50 +0000 (14:59 -0600)]
t/re/regex_sets.t: Generalize for non-ASCII platforms
Karl Williamson [Tue, 19 Mar 2013 21:27:31 +0000 (15:27 -0600)]
t/op/quotemeta.t: EBCDIC fixes
Karl Williamson [Tue, 19 Mar 2013 17:32:55 +0000 (11:32 -0600)]
t/re/fold_grind.t: Generalize for non-ASCII platforms
Karl Williamson [Tue, 19 Mar 2013 17:20:24 +0000 (11:20 -0600)]
t/uni/class.t: Generalize to work on non-ASCII platforms
Karl Williamson [Tue, 19 Mar 2013 17:01:57 +0000 (11:01 -0600)]
feature/unicode_strings.t: Generalize to work on non-ASCII platforms
Karl Williamson [Sun, 2 Nov 2014 03:10:48 +0000 (21:10 -0600)]
lib/dumpvar.pl: Generalize for non-ASCII platforms
Karl Williamson [Fri, 15 Mar 2013 17:57:24 +0000 (11:57 -0600)]
re/charset.t: Generalize for non-ASCII platforms
This converts the hard-coded character numbers to native, so will work
on any platform. It also adds some tests, and improves some test names
Karl Williamson [Wed, 13 Mar 2013 22:17:39 +0000 (16:17 -0600)]
APItest/t/handy.t: Generalize for non-ASCII platforms
Karl Williamson [Mon, 11 Mar 2013 04:17:31 +0000 (22:17 -0600)]
t/base/lex.t: Generalize for EBCDIC platforms
\xE2 is 'S' in EBCDIC, and so is ASCII there. We need something that is
an alpha on both platforms that isn't ASCII. \xDF works.
Also, 65 may not be "A" on a non-ASCII platform, so use 65 instead of
'A' where 65 is wanted
Karl Williamson [Wed, 26 Jun 2013 22:06:47 +0000 (16:06 -0600)]
lib/Unicode/UCD.t: White-space only
Properly indent within a block
Steve Hay [Thu, 5 Mar 2015 23:00:35 +0000 (23:00 +0000)]
perldelta for previous commit (threads upgrade)
Jerry D. Hedden [Thu, 5 Mar 2015 05:28:51 +0000 (00:28 -0500)]
Upgrade to threads 1.98
Karl Williamson [Fri, 21 Nov 2014 16:22:53 +0000 (09:22 -0700)]
lib/Config.t: Skip tests that are invalid on os390
Karl Williamson [Wed, 29 Oct 2014 16:08:46 +0000 (10:08 -0600)]
t/comp/require.t: No longer needed to skip on EBCDIC
Karl Williamson [Mon, 9 Jun 2014 18:30:23 +0000 (12:30 -0600)]
t/uni/write.t: Don't skip EBCDIC tests
This now passes on os390
Karl Williamson [Wed, 26 Jun 2013 21:28:41 +0000 (15:28 -0600)]
t/lib/warnings/utf8: Update skip message
It is not that EBCDIC platforms don't check for malformedness; it is
that a different message gets generated on them, and it's too much
bother to figure it out for each such platform.
Karl Williamson [Mon, 2 Mar 2015 22:03:08 +0000 (15:03 -0700)]
t/test.pl: EBCDIC, utf8 fixes
This changes the display of all low-ordinal EBCDIC controls to octal,
instead of just the ones that correspond to the ASCII C0 controls.
It also makes sure that [:print:] thinks all non-ASCII are
non-printable.
Karl Williamson [Mon, 2 Mar 2015 22:00:25 +0000 (15:00 -0700)]
lib/Unicode/UCD.t: Fixes to work on EBCDIC
These were broken by the upgrade of Unicode::UCD to version 0.61
Karl Williamson [Thu, 5 Mar 2015 19:08:48 +0000 (12:08 -0700)]
lib/open.t: Fix skip count
Karl Williamson [Thu, 5 Mar 2015 17:34:08 +0000 (10:34 -0700)]
Data::Dumper/t/dumper.t: Fix comments
Rafael Garcia-Suarez [Thu, 5 Mar 2015 16:37:33 +0000 (17:37 +0100)]
readline should behave as <>, not <<>>
This fixes bug #123990. The cause was that one promotion to
OP_READLINE was not cleaning the special flag on that op.
David Mitchell [Thu, 5 Mar 2015 11:44:37 +0000 (11:44 +0000)]
sprinkle NOTREACHED and FALLTHROUGH
Coverity complains bitterly about many switch statements in lots of files.
Many of these are of the form:
case FOO:
...
goto baz;
case BAR:
....
and something as smart as Coverity shouldn't really be complaining about
a missing 'break' when the last statement of the previous branch is an
unconditional goto/return/continue. But I've shoved in a bunch of
'NOTREACHED' to hopefully shut it up.
Missing 'FALLTHROUGH' comments were more reasonable, and I've added them
where appropriate.
The only confusing one was cx_dup(), where the various CXt_LOOP_ branches
all fell through to the next one, and it took a while to figure out that
those weren't bugs.
David Mitchell [Thu, 5 Mar 2015 11:01:01 +0000 (11:01 +0000)]
pp.c: remove unneeded SPAGAIN's
Spotted by Coverity.
I've left the SPAGAIN but commented out, in case future changes
to the code make them needed again.
David Mitchell [Thu, 5 Mar 2015 10:28:28 +0000 (10:28 +0000)]
document why S_maybe_multideref is Coverity-safe
David Mitchell [Wed, 4 Mar 2015 14:49:22 +0000 (14:49 +0000)]
followup for MEM_WRAP_CHECK
On my previous commit, I failed to save the edited handy.h file where
I had edited the comments, so they were missed from the commit.
David Mitchell [Wed, 4 Mar 2015 14:22:27 +0000 (14:22 +0000)]
further tweak MEM_WRAP_CHECK()
My recent mods to make it often constant-fold at time-time were
generating Coverity warnings when the expression happened to equal
(cond ? 1 : 1).
Jarkko Hietaniemi [Wed, 4 Mar 2015 12:37:33 +0000 (07:37 -0500)]
Consistently use NOT_REACHED; /* NOTREACHED */
Both needed: the macro is for compilers, the comment for static checkers.
(This doesn't address whether each spot is correct and necessary.)
Steve Hay [Wed, 4 Mar 2015 08:43:41 +0000 (08:43 +0000)]
Upgrade Text-Balanced from version 2.02 to 2.03
This incorporates the blead customizations, with no other changes except
the version bump.
Steve Hay [Wed, 4 Mar 2015 08:39:17 +0000 (08:39 +0000)]
threads is no longer customized, as of commit
c0ff91434b
Steve Hay [Wed, 4 Mar 2015 08:14:20 +0000 (08:14 +0000)]
threads is no longer customized, as of commit
c0ff91434b
Jerry D. Hedden [Tue, 3 Mar 2015 23:30:54 +0000 (18:30 -0500)]
Upgrade to threads 1.97
Committer actions: (1) regen for t/porting/customized.dat.
(2) perldelta for update to threads.
Jarkko Hietaniemi [Wed, 4 Mar 2015 00:15:25 +0000 (19:15 -0500)]
IRIX: floating point: do not flush to zero
Fix for [perl #123767] IRIX64 blead (
ddce084a) opbasic/arith.t failure
Use the PERL_SYS_FPU_INIT to initialize the FPU flags appropriately.
David Mitchell [Thu, 26 Feb 2015 18:42:16 +0000 (18:42 +0000)]
Make MEM_WRAP_CHECK more compile-time
MEM_WRAP_CHECK(n,t) checks whether n * sizeof(t) exceeds the
memory size, and so is likely to wrap.
When the type of n is small (e.g. a U8), you used to get compiler warnings
about a comparison always being true. This was avoided by adding 0.0. Now
Coverity complains that you're doing a floating-point comparison with the
results of an integer division.
Instead of adding 0.0, instead add some more compile-time checks
that will cause the runtime check to be skipped when the maximum value
of n (as determined by sizeof(n)) is a lot less than memory size.
On my 64-bit system this also pleasingly makes the executable 8384 bytes
smaller, implying that in many cases, the run-time check is now being
skipped.
Jarkko Hietaniemi [Tue, 3 Mar 2015 01:46:53 +0000 (20:46 -0500)]
Tru64: living dangerously, enabling the -c99 flag for cc
16 years after C99 and 2 years after Tru64 EOL.
James E Keenan [Thu, 26 Feb 2015 14:02:19 +0000 (09:02 -0500)]
Remove unused variable $DEBUGGING.
It is unused and there do not appear to be any tests to be run for -DDEBUGGING
in this file.
Sullivan Beck [Mon, 2 Mar 2015 23:37:20 +0000 (18:37 -0500)]
Bump Locale-Codes from 3.33 to 3.34
The core modules Locale::Country, Locale::Language, and Locale::Currency (all
part of the Locale-Codes distribution) should be updated on a regular basis.
They contain "codes" from various Internet standards which change over time.
David Mitchell [Mon, 2 Mar 2015 15:50:09 +0000 (15:50 +0000)]
regen uconfig.h
Needed after
0cd93aca4e274765.
Jarkko Hietaniemi [Mon, 2 Mar 2015 12:00:33 +0000 (07:00 -0500)]
Configure: add note about nvsize sometimes lying
Lying in case you expect that many bytes being used for NV.
Father Chrysostomos [Mon, 2 Mar 2015 06:42:04 +0000 (22:42 -0800)]
[perl #123960] sv.c: Fix gp_free -Do output
We try to print ‘gp_free clearing PL_stashcache for '%"HEKf"'’, but
without checking that we actually have a hek. This has been this way
since
103f5a36127.
Tony Cook [Mon, 2 Mar 2015 06:10:15 +0000 (17:10 +1100)]
perldelta for
b6bba8869651
Tony Cook [Mon, 2 Mar 2015 06:03:12 +0000 (17:03 +1100)]
[perl #123947] deparse $#_ as $#_ not as $#{_}
8c2e27d changed B::Deparse to deparse $#{1} as that instead of as $#1
which is a syntax error, but also changed $#_ to deparse as $#{_}
which isn't necessary, and broke GFUJI/macro-0.06.tar.gz
This could be considered a bug in macro.pm, but since we don't need to
deparse $#_ as $#{_} let's not do that.
Father Chrysostomos [Mon, 2 Mar 2015 02:20:53 +0000 (18:20 -0800)]
[perl #123849] sv.c: Fix sv_clear -Do output
We try to print ‘sv_clear clearing PL_stashcache for '%"SVf"'’, but
the sv here is a hash, and we can’t do SvPV on it. We need the hash’s
name. This has been this way since
103f5a36127.
Father Chrysostomos [Mon, 2 Mar 2015 02:11:23 +0000 (18:11 -0800)]
[perl #123955] Fix assert fail with 0 s/// in quotes
This is related to bug #123617 and is a follow-up to
eabab8bcc.
This code:
"@0{0s 000";eval"$"
begins tokenisation as follows:
stringify ( join ( $ " , @ 0 { 0 subst
When seeing the subst after the 0, the parser discards many tokens and
we end up at the ; outside the quotes.
Since PL_lex_stuff (the temporary spot for storing the contents of a
quote-like operator) is localised as of
eabab8bcc, we end up with just
PL_sublex_info.repl (the temporary spot for storing the replacement
part) set. Since it is still set when we get to the next double-
quote, it is treated as a two-part quote-like operator, like y or s.
That can’t happen, and we have assertions to make sure of it.
We need to localise PL_sublex_info.repl as well, so it gets freed
properly when scopes are popped after an error.
Father Chrysostomos [Sun, 1 Mar 2015 22:01:27 +0000 (14:01 -0800)]
[perl #123848] Allow OPpTARGET_MY on qr
// in the scope of lexical $_ gets the OPpTARGET_MY flag set. If it
is used as an operand to smartmatch, it gets converted to a qr op.
There is no need to turn off the flag, since it is harmless at run
time. But we need to allow the flag on this op type to avoid asser-
tion failures when it is freed.
Father Chrysostomos [Sun, 1 Mar 2015 21:27:39 +0000 (13:27 -0800)]
op.c apidoc typo
Father Chrysostomos [Sun, 1 Mar 2015 21:26:52 +0000 (13:26 -0800)]
[perl #123821] Fix assert fail with \(&$0)=0
If the operand of a reference constructor that is assigned to is not
valid, then S_lvref (in op.c) queues an error and does not bother to
convert the op into an lvref op. This causes problems, since we have
already set lvref-specific flags on the op in the expectation that its
type will change. We get assertion failures later when freeing the
op, because of the invalid flags.
Instead of bailing out when we have an error like this, just continue
processing it, as we do in many other places.
Father Chrysostomos [Sun, 1 Mar 2015 21:16:36 +0000 (13:16 -0800)]
[perl #123817] Assert fail with attr in anon hash
These two one-liners were failing an assertion:
$y = { my $x : m }
$y = [ (my $x : m) ]
The lexical with an attribute was turning into an op of type ‘list’
with the OPpLVAL_INTRO flag set. Many op constructors convert an
operand that is of type ‘list’ into the correct type, instead of
creating a new op to wrap around it. Other ops were not expecting the
flag to be set, causing an assertion failure. The easiest solution is
to turn off the flag in op_convert_list, which is where the op type
changes. (I tried modifying other op types to accept the flag, but
that caused problems for B::Deparse, which inquires of B::Op_private
which ops can legitimately take that flag.)
Father Chrysostomos [Sun, 1 Mar 2015 19:43:11 +0000 (11:43 -0800)]
op.c: Remove redundant op_lvalue call
Refgen ops already call op_lvalue on their operands. Calling it
beforehand causes it to be called twice, which is harmless, but
wasteful.
Father Chrysostomos [Sun, 1 Mar 2015 19:29:10 +0000 (11:29 -0800)]
[perl #123802] Assertion failure with "\L\L"
This is a follow-up to
f4460c6f7a. The check to see whether we are
in a quote-like operator needs to come before the call to sublex_done,
as sublex_done is just as problematic as doing SvIVX on a PV. (See
479ae48e22f for details on why.) Checking the type of PL_linestr is
not a reliable way to see whether we are in a quote-like op, so use
PL_in_what instead.
Father Chrysostomos [Sun, 1 Mar 2015 19:20:47 +0000 (11:20 -0800)]
t/op/lex.t: Correct bug num
Father Chrysostomos [Sun, 1 Mar 2015 06:31:25 +0000 (22:31 -0800)]
[perl #123763] Clear target on my $_; split
If a lexical $_ is in scope, then the first argument to split, which
starts out as a match op, will get the pad offset of $_ as its target,
since that’s how implicit lexical $_=~ usually works.
ck_split changes that first argument to a pushre op. The target
was not being cleared. That did not cause any problems, until
v5.21.4-408-gfd017c0, which optimised lexical @array = split to write
to split @array directly, by storing making lexical array’s pad offset
the pushre’s target.
You can see the obvious conflict there. We end up trying to split to
$_, which is not an array. On a debugging build, you get an assertion
failure when trying to extend $_. Make the split list long enough,
and you get a crash on non-debugging builds.
debugging$ ./miniperl -e 'my $_; split'
Use of my $_ is experimental at -e line 1.
Assertion failed: (SvTYPE(av) == SVt_PVAV), function Perl_av_extend, file av.c, line 70.
non-debugging$ ./miniperl -e 'my $_; split //, "a"x100000'
Use of my $_ is experimental at -e line 1.
Segmentation fault: 11
Father Chrysostomos [Sun, 1 Mar 2015 06:12:30 +0000 (22:12 -0800)]
Test for split after my $_ [perl #123763]
Hugo van der Sanden [Sun, 1 Mar 2015 02:58:58 +0000 (18:58 -0800)]
[perl #123763] pushre needs to permit OPpTARGET_MY
This case fails an assertion:
my($_);0=split
because a pushre op is not expected to have that flag. A pushre op
starts out as a match op (which may indeed have the flag set), as is
subsequently converted to a pushre op.
[The commit message was written by the committer.]
Father Chrysostomos [Sun, 1 Mar 2015 01:41:36 +0000 (17:41 -0800)]
[perl #123802] Assertion failure with /$0{}/
If the parser (perly.c) pops tokens when encountering a syntax error,
it can pop inner lexing scopes (which handle the contents of quote-
like operators).
If there is a pending token on the pending token stack, then the cur-
rent lexing state is LEX_KNOWNEXT. It usually gets set to a pending
value stored in PL_lex_defer when the last pending token is emitted.
If scopes are exited when there is a pending token, then the state is
reset, since it is localised, even thought we still have a token pend-
ing. We have code to account for that and still emit the pending
token. (See
7aa8cb0dec1.) But the pending lexing state is still
used after the pending token is emitted. So we can end up with
LEX_INTERPEND when there is no inner lexing scope.
LEX_INTERPEND will cause sublex_done (in toke.c) to be called, which
does a LEAVE. If the current scope does not belong to it, we end up
exiting a scope set up by the parser, which frees the parser stack
(via SAVEDESTRUCTOR_X and clear_yystack in perly.c). The parser is
still using the stack, which holds reference counts on the CV in
PL_compcv, so reference counts get screwed up.
We need to check whether we have a proper lexing scope set up if the
lexing state is LEX_INTERPEND.
This is a follow-up to
f4460c6f7a, which was a similar bug, but
occurred with LEX_INTERPCONCAT, rather than LEX_INTERPEND.
Jarkko Hietaniemi [Fri, 27 Feb 2015 21:37:57 +0000 (16:37 -0500)]
infnan: NV_MANT_REAL_DIG is unused
(now that the nan payload code is gone)
David Mitchell [Fri, 27 Feb 2015 21:06:13 +0000 (21:06 +0000)]
fix to "fix op leak caused by OP_MULTIDEREF"
Update to
b7613b8a45c70113.
Hugo pointed out that my for(; p; p = OpSIBLING(p)) op_free(p) loop
relied on p->op_sibling still being valid after op_free(p).
Karl Williamson [Fri, 27 Feb 2015 18:57:03 +0000 (11:57 -0700)]
Revert "perlpod and spec: s/Latin-1/CP-1252/"
This reverts commit
e2bb786192adfa315ea974b5f630d7040aa6f6ac.
These doc changes apply to Pod::Simple version 3.30, which, alas, was
created too late to be included in Perl v5.22, so this change to the
docs in blead is premature.
I plan to apply the patch in 5.23.
Ricardo Signes [Fri, 27 Feb 2015 03:35:52 +0000 (22:35 -0500)]
perlpod and spec: s/Latin-1/CP-1252/