This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Aristotle Pagaltzis [Sun, 18 Sep 2016 09:53:20 +0000 (11:53 +0200)]
perlfunc: re-document old split() @_ side effect
Lukas Mai [Sun, 18 Sep 2016 07:50:16 +0000 (09:50 +0200)]
perlsub: scalar split no longer clobbers @_ (RT #129297)
Karl Williamson [Sun, 18 Sep 2016 03:07:29 +0000 (21:07 -0600)]
perldelta for new Unicode-handling function.
Karl Williamson [Thu, 15 Sep 2016 01:49:52 +0000 (19:49 -0600)]
perlapi: Clarify docs for some is_utf8_foo functions
Karl Williamson [Thu, 15 Sep 2016 00:54:23 +0000 (18:54 -0600)]
Add isUTF8_CHAR_flags() macro
This is like the previous 2 commits, but the macro takes a flags
parameter so any combination of the disallowed flags may be used. The
others, along with the original isUTF8_CHAR(), are the most commonly
desired strictures, and use an implementation of a, hopefully, inlined
trie for speed. This is for generality and the major portion of its
implementation isn't inlined.
Karl Williamson [Mon, 12 Sep 2016 22:52:41 +0000 (16:52 -0600)]
Add macro for Unicode Corregindum #9 strict
This macro follows Unicode Corrigendum #9 to allow non-character code
points. These are still discouraged but not completely forbidden.
It's best for code that isn't intended to operate on arbitrary other
code text to use the original definition, but code that does things,
such as source code control, should change to use this definition if it
wants to be Unicode-strict.
Perl can't adopt C9 wholesale, as it might create security holes in
existing applications that rely on Perl keeping non-chars out.
Karl Williamson [Mon, 12 Sep 2016 19:38:22 +0000 (13:38 -0600)]
Add macro for determining if UTF-8 is Unicode-strict
Karl Williamson [Mon, 12 Sep 2016 20:30:15 +0000 (14:30 -0600)]
perlapi: Clarify isUTF8_CHAR()
Karl Williamson [Wed, 14 Sep 2016 23:09:51 +0000 (17:09 -0600)]
inline.h: Add 'const's; avoid hiding outer variable
This changes some formal parameters to be const, and avoids reusing the
same variable name within an inner block, to avoid confusion
Karl Williamson [Thu, 8 Sep 2016 17:34:15 +0000 (11:34 -0600)]
Add tests for is_valid_partial_utf8_char_flags()
Karl Williamson [Mon, 12 Sep 2016 04:18:57 +0000 (22:18 -0600)]
Add is_utf8_valid_partial_char_flags()
This is a generalization of is_utf8_valid_partial_char to allow the
caller to automatically exclude things such as surrogates.
Karl Williamson [Sun, 11 Sep 2016 15:40:37 +0000 (09:40 -0600)]
perlapi: Reword description of is_utf8_valid_partial_char
Karl Williamson [Sun, 11 Sep 2016 04:27:37 +0000 (22:27 -0600)]
Fix off-by-one error in is_utf8_valid_partial_char()
Karl Williamson [Sun, 11 Sep 2016 04:24:48 +0000 (22:24 -0600)]
handy.h: Comment memEQs and memNEs
Karl Williamson [Sun, 11 Sep 2016 04:18:59 +0000 (22:18 -0600)]
utf8.c: Add some UNLIKELYs
Karl Williamson [Sun, 11 Sep 2016 04:18:16 +0000 (22:18 -0600)]
utf8.h: Add comment, white-space changes
Karl Williamson [Sun, 11 Sep 2016 04:09:44 +0000 (22:09 -0600)]
Enhance and rename is_utf8_char_slow()
This changes the name of this helper function and adds a parameter and
functionality to allow it to exclude problematic classes of code
points, the same ones excludeable by utf8n_to_uvchar(), like surrogates
or non-character code points.
Karl Williamson [Thu, 8 Sep 2016 04:22:01 +0000 (22:22 -0600)]
APItest/t/utf8.t: Add tests
These fill in gaps in current testing. In particular all the overlong
UTF-8 possible edge cases are now tested.
Karl Williamson [Thu, 8 Sep 2016 04:14:38 +0000 (22:14 -0600)]
APItest/utf8.t: Some clean up
This adds some information to test names, does some white-space
alignments, changes one test to stress things slightly more, and adds a
'use bytes' because in some cases the desired byte-oriented output was
not showing up.
Karl Williamson [Mon, 5 Sep 2016 03:32:08 +0000 (21:32 -0600)]
Test isUTF8_CHAR()
Karl Williamson [Sun, 11 Sep 2016 04:19:42 +0000 (22:19 -0600)]
lib/warnings/utf8: Reinstate warning test
I removed this in
35f8c9bd0ff4f298f8bc09ae9848a14a9667a95a, thinking the
warning was no longer being raised. But in fact, it was showing a bug,
now fixed by the previous commit.
Karl Williamson [Sun, 11 Sep 2016 03:15:04 +0000 (21:15 -0600)]
Revamp overlong handling in is_utf8_char_slow, fixing a bug
This combines EBCDIC and ASCII branches as much as possible, and fixes a
bug that showed up only on EBCDIC platforms, and 64-bit ASCII ones for
the highest overlong, where it could erroneously conclude that a
sequence was an overlong.
Tests are coming in a future commit.
.
Karl Williamson [Sun, 11 Sep 2016 03:06:39 +0000 (21:06 -0600)]
utf8.c: Fix typo in comment, add some comments
Karl Williamson [Sat, 10 Sep 2016 15:00:03 +0000 (09:00 -0600)]
utf8.c: Extract duplicate code to common fcn
Actually the code isn't quite duplicate, but should be because one
instance is wrong. This failure would only show up on EBCDIC platforms.
Tests are coming in a future commit.
Karl Williamson [Sat, 10 Sep 2016 14:54:36 +0000 (08:54 -0600)]
handy.h: Add memLT, memLE, memGT, memGE
These correspond to strLT, etc. I am deferring documenting them in case
this turns out to be a bad idea for some reason.
Karl Williamson [Sat, 10 Sep 2016 14:46:18 +0000 (08:46 -0600)]
Unconditionally define memcmp() if not sane
Prior to this commit, if there was a #define for memcmp that invoked a
version that Configure deemed to not be sufficient for normal use, it
was retained, so that perl used the defective version. This apparently
hasn't been a problem in the field, but I realized the potential issue
doing code reading, and am correcting it.
Karl Williamson [Sat, 3 Sep 2016 20:12:27 +0000 (14:12 -0600)]
isUTF8_CHAR(): Bring UTF-EBCDIC to parity with ASCII
This changes the macro isUTF8_CHAR to have the same number of code
points built-in for EBCDIC as ASCII. This obsoletes the
IS_UTF8_CHAR_FAST macro, which is removed.
Previously, the code generated by regen/regcharclass.pl for ASCII
platforms was hand copied into utf8.h, and LIKELY's manually added, then
the generating code was commented out. Now this has been done with
EBCDIC platforms as well. This makes regenerating regcharclass.h
faster.
The copied macro in utf8.h is moved by this commit to within the main
code section for non-EBCDIC compiles, cutting the number of #ifdef's
down, and the comments about it are changed somewhat.
Karl Williamson [Sat, 3 Sep 2016 18:15:29 +0000 (12:15 -0600)]
regen/regcharclass.pl: surrogates are code points
They are not "characters"
Karl Williamson [Sat, 3 Sep 2016 22:13:15 +0000 (16:13 -0600)]
Add IS_UTF8_INVARIANT and IS_UVCHR_INVARIANT to API
Karl Williamson [Thu, 8 Sep 2016 04:03:21 +0000 (22:03 -0600)]
utfebcdic.h: Fix typo in comment
Karl Williamson [Wed, 14 Sep 2016 22:05:35 +0000 (16:05 -0600)]
Add #defines for XS code for Unicode Corregindum 9
These are convenience macros.
Karl Williamson [Wed, 14 Sep 2016 22:02:50 +0000 (16:02 -0600)]
perlapi: Clarify utf8n_to_uvchr entry
Karl Williamson [Wed, 14 Sep 2016 21:57:34 +0000 (15:57 -0600)]
perlunicode: Fix typo
Karl Williamson [Tue, 13 Sep 2016 22:40:44 +0000 (16:40 -0600)]
append_utf8_from_native_byte: Add parens for clarity
I can never remember the precedence of dereference and ++.
E. Choroba [Sat, 17 Sep 2016 20:05:15 +0000 (22:05 +0200)]
Fix English in perlfunc
Father Chrysostomos [Sat, 17 Sep 2016 18:06:55 +0000 (11:06 -0700)]
Add regexp_nonull.t
for testing the regular expression engine with strings that
lack a trailing null byte.
E. Choroba [Fri, 16 Sep 2016 08:33:53 +0000 (01:33 -0700)]
Nested single quotes in documentation example
Documentation of the Unicode Bug contains an example that nests single
quotes in a shell. Most shells can't do that. Patch attached.
Signed-off-by: Abigail <abigail@abigail.be>
Lukas Mai [Fri, 16 Sep 2016 11:02:16 +0000 (13:02 +0200)]
restore comment mangled by
27da23d53ccce6
Sawyer X [Wed, 14 Sep 2016 21:29:46 +0000 (23:29 +0200)]
Paying respect to Jon AVENJ Portnoy
Steve Hay [Wed, 14 Sep 2016 13:08:37 +0000 (14:08 +0100)]
Upgrade Test::Simple from version 1.302052 to 1.302056
Father Chrysostomos [Wed, 14 Sep 2016 06:05:43 +0000 (23:05 -0700)]
Increase $XS::APItest::VERSION to 0.84
Father Chrysostomos [Wed, 14 Sep 2016 06:01:58 +0000 (23:01 -0700)]
perldelta for #129164 /
92b69f650
Father Chrysostomos [Wed, 14 Sep 2016 06:00:56 +0000 (23:00 -0700)]
perldelta for #129090 /
6da13066b6
Father Chrysostomos [Wed, 14 Sep 2016 05:59:18 +0000 (22:59 -0700)]
perldelta for #47047 /
1de22db27a
Father Chrysostomos [Wed, 14 Sep 2016 05:57:22 +0000 (22:57 -0700)]
perldelta for #129196 /
9bde56224
Father Chrysostomos [Wed, 14 Sep 2016 05:54:49 +0000 (22:54 -0700)]
Father Chrysostomos [Wed, 14 Sep 2016 05:50:30 +0000 (22:50 -0700)]
perldelta for #125679 /
2b6a5bfb1
Father Chrysostomos [Wed, 14 Sep 2016 05:48:29 +0000 (22:48 -0700)]
perldelta for #107726 /
8bc40f3a4e
Father Chrysostomos [Wed, 14 Sep 2016 05:46:43 +0000 (22:46 -0700)]
Father Chrysostomos [Wed, 14 Sep 2016 05:38:59 +0000 (22:38 -0700)]
[perl #129267] Test for gv_fetchmethod buffer overrun
Yves Orton [Tue, 13 Sep 2016 21:14:49 +0000 (23:14 +0200)]
fix #129267: rework gv_fetchmethod_pvn_flags separator parsing
With previous code we could overrun the end of the name when
the last char in the string was a colon. This reworks the code
so it is more clear what is going on, and so it more similar
to other code that also parses out package separaters in gv.c.
This is a rework of the reverted patches:
243ca72 rename "nend" name_cursor in Perl_gv_fetchmethod_pvn_flags
b053c93 fix: [perl #129267] Possible string overrun with invalid len in gv.c
Yves Orton [Tue, 13 Sep 2016 21:10:48 +0000 (23:10 +0200)]
clean up gv_fetchmethod_pvn_flags: rename nsplit to last_separator
nsplit if set points at the first char of the last separator
in name, so rename it so it is more comprehensible what it means.
Yves Orton [Tue, 13 Sep 2016 21:06:58 +0000 (23:06 +0200)]
clean up gv_fetchmethod_pvn_flags: move origname init to function start
so it is more obvious that it is a constant copy of the
original name.
Yves Orton [Tue, 13 Sep 2016 21:06:07 +0000 (23:06 +0200)]
clean up gv_fetchmethod_pvn_flags: introduce name_end
nend is used for too many things, this replaces various
uses of nend with name_end, which is constant.
this is a first step to fixing [perl #129267], which shouldnt
change any behavior
Yves Orton [Tue, 13 Sep 2016 19:53:48 +0000 (21:53 +0200)]
Revert "fix: [perl #129267] Possible string overrun with invalid len in gv.c"
This reverts commit
b053c9385f29680bc812db3da64ce5c4d45d0cf7.
Yves Orton [Tue, 13 Sep 2016 19:53:36 +0000 (21:53 +0200)]
Revert "rename "nend" name_cursor in Perl_gv_fetchmethod_pvn_flags"
This reverts commit
243ca72243bd234ab5ef1818a32b9c6dd68f777c.
Yves Orton [Tue, 13 Sep 2016 19:21:08 +0000 (21:21 +0200)]
rename "nend" name_cursor in Perl_gv_fetchmethod_pvn_flags
nend is a crappy name, and we use name_cursor in other
contexts that do similar things. consistency++
Yves Orton [Tue, 13 Sep 2016 19:19:53 +0000 (21:19 +0200)]
fix: [perl #129267] Possible string overrun with invalid len in gv.c
Perl_gv_fetchmethod_pvn_flags contains various subtle logic bugs related to parsing
fully qualified method names. In particular if you feed the function a string
which ends with a single colon it will end up accessing memory past the end
of the string. In addition when checking for the second colon in a fully qualified
name we could potentially access memory we dont own, and certainly access memory
not part of the string
Father Chrysostomos [Mon, 12 Sep 2016 04:29:56 +0000 (21:29 -0700)]
[perl #129164] Crash with splice
This fixes #129166 and #129167 as well.
splice needs to take into account that arrays can hold NULLs and
return &PL_sv_undef in those cases where it would have returned a
NULL element.
Father Chrysostomos [Sat, 10 Sep 2016 14:27:43 +0000 (07:27 -0700)]
pad.c:pad_fixup_inner_anons: Add assertions
These would have made it easier to track down the bug fixed by
the previous commit.
Father Chrysostomos [Sat, 10 Sep 2016 14:24:41 +0000 (07:24 -0700)]
[perl #129090] Crash with sub c{sub c}
The crash in the bug report was caused by v5.21.7-215-g307cbb9, which
added the code in question in order to fix another crash. It exposed
an existing problem caused by v5.21.6-386-ga70f21d.
Some background:
When parsing ‘sub c{’, perl creates a new CV. The CV gets installed
in the symbol table when the final ‘}’ is reached.
If there is already an undefined CV at that point, it gets reused and
the contents of the new CV are transferred to the existing one. That
means that any subs defined within now have their CvOUTSIDE pointer
pointing to a stub that is about to be freed, so pad_fix_inner_anons
gets called to update those CvOUTSIDE pointers accordingly.
Formats were not getting their CvOUTSIDE pointers updated, so commit
v5.11.2-160-g421f30e added formats to the containing sub’s pad to
allow the CvOUTSIDE fix-up to apply to them, too.
That caused a crash, as explained in the commit message for
v5.17.1-213-ge09ac07, that the cited commit fixed by putting a weak RV
in the pad, not a direct pointer to the format, and giving the format
a strong CvOUTSIDE pointer.
Problem:
Commit v5.21.6-386-ga70f21d¹ fixed a problem with named subs not cor-
rectly referencing outer state vars (because of bad CvOUTSIDE point-
ers hiding the lexical vars). It did so by extending the mechanism
already used for formats to apply to named subs as well.
The one mistake that that commit made was to add a pad entry for the
sub even if it was just a stub declaration. That caused a sub with a
stub declaration inside it to a loop if they have the same name:
$ ./perl -Ilib -le '%c; sub c { sub c; eval q|$x| } c'
This happens because the ‘sub c;’ inserts a stub into the symbol table
and inserts a reference to that stub also into the pad of the sub cur-
rently being compiled. The final CvOUTSIDE fix-up sets the CvOUTSIDE
of the erstwhile stub (that has just become the newly defined sub) to
the sub itself. (The %c in the example is necessary to vivify the
typeglob; otherwise perl will cheat and the stub declaration won’t
actually create a CV.)
In addition, a crash can occur in this one-liner (reduced from the
original bug report):
$ ./miniperl -e '$a="a$a";my sub b;%c;sub c{sub b;sub c}'
The fix-up code iterates through the pad backwards, starting from the
last entry, so the CvOUTSIDE(c) -> c loop is set up before the lexical
&b entry is encountered. When it is encountered, CvOUTSIDE pointers
are followed to find the outer pad where b is defined, but the
PARENT_PAD_INDEX stored in the &b padname is wrong for c’s own pad,
which is what we end up looking at due to the CvOUTSIDE loop. The
‘outer’ entry may then turn out be an RV, not a CV, or just about any-
thing. Crashes ensue.
Another, perhaps clearer, way to make it crash is to give the lexical
sub pad entry an outer offset far beyond the end of the directly-
enclosing sub:
$ ./perl -Ilib -lE '%c; my ($a,$b,$c,$d,$e,$f,$g,$h,$i,$j,$k,$l,$m,$n,$o,$p,$q,$r,$s,$t,$u); state sub b; sub c { sub b {} sub c }'
Segmentation fault: 11
Solution:
The fix is to stop package stubs from creating pad entries.
¹ Which I found by bisecting like this, in case anyone is interested:
../perl.git/Porting/bisect.pl --expect-fail --start=v5.20.0 --end=v5.22.0 -e '%c; sub c{sub c}; use B; print B::svref_2object(\&c)->PADLIST->ARRAYelt(0)->ARRAYelt(1)->PV'
Father Chrysostomos [Fri, 9 Sep 2016 05:17:34 +0000 (22:17 -0700)]
op.c: Avoid string cmp when unnecessary
Father Chrysostomos [Thu, 8 Sep 2016 21:25:09 +0000 (14:25 -0700)]
toke.c: Correct comment
Aaron Crane [Sat, 10 Sep 2016 09:44:34 +0000 (10:44 +0100)]
Swap release managers for 5.25.6 and 5.25.7
Craig A. Berry [Fri, 9 Sep 2016 19:05:38 +0000 (14:05 -0500)]
Force a newline in new attrs.t test.
VMS is going to give you newline at EOF willy nilly, so make it
explicit and match expectations everywhere.
Yaroslav Kuzmin [Fri, 2 Sep 2016 20:27:21 +0000 (14:27 -0600)]
perlio.c: Fix z/OS problem
See thread starting at
<
6E741FE88A7C4E428528A0AAAB94D735B65C65@DFLE10.ent.ti.com>
(The commiter made some white space changes to the patch)
David Mitchell [Thu, 8 Sep 2016 07:55:45 +0000 (08:55 +0100)]
Revert "tmp fix for Bleadperl breaks Variable-Magic"
This reverts commit
8d168aaa014262c7f93944b76b84de99af3c5513.
Variable-Magic 0.60 has been released, so this temp workaround is
no longer required.
James E Keenan [Fri, 5 Aug 2016 23:13:15 +0000 (19:13 -0400)]
Add additional tests for DirHandle to improve coverage.
Add descriptions to all tests. Revise per suggestions by Tony Cook.
For: RT #128856
David Mitchell [Wed, 7 Sep 2016 19:57:01 +0000 (20:57 +0100)]
rename S_delimcpy() to S_delimcpy_intern()
Its a bit confusing having both S_delimcpy() and Perl_delimcpy()
functions.
David Mitchell [Fri, 26 Aug 2016 11:10:58 +0000 (12:10 +0100)]
add some code comments for the users of delimcpy()
While fixing delimcpy(), I found that it wasn't always clear what its
callers did, so I've added some extra code comments.
also add a balancing '}' in a comment block to help editors that
jump between matching brackets.
David Mitchell [Thu, 25 Aug 2016 16:48:34 +0000 (17:48 +0100)]
Perl_delimcpy(): handle backslash as last char
[perl #129064] heap-buffer-overflow S_scan_heredoc
[perl #129176] Conditional jump depends on uninitialized values in
S_scan_heredoc
Perl_delimcpy() is supposed to copy a delimited string to another buffer;
it handles \-<delimiter> escapes, but if the backslash is the last
character in the src buffer, it could overrun the end of the buffer
slightly.
Also document a bit better what this function is supposed to do.
Lukas Mai [Wed, 7 Sep 2016 17:30:37 +0000 (19:30 +0200)]
perlobj: s/Deferencing/Dereferencing/
David Mitchell [Wed, 7 Sep 2016 08:30:26 +0000 (09:30 +0100)]
fix errror handling for ':attr(foo' with no ')'
When the parameter of an attribute has no matching closing ')', there
are several issues with the way the error is handled.
First, the code currently tries to set PL_bufptr back to the position
of the opening '(' for a better error message. However, since the error
will have been discovered only at EOF after all the remaining lines have
been read and discarded, the buffer no longer contains ':attr(...'.
So the error message includes a spurious \0 followed by possibly some
random chunk of characters from the last line read in.
Worse, if the input buffer gets realloced while perl searches for the ')',
then PL_bufptr is reset to point into the freed buffer. [perl #129086].
It does yyerror() rather than croak(), so further error messages appear,
even though we're at EOF and no further parsing can occur. Similar cases
such as no matching closing string delimiter all croak instead.
It resets cop_line to zero *before* printing the error, so the line number
of the error is erroneously reported as zero.
This commit fixes all these issues by handling the error more similarly
to the way unterminated strings are handled. In particular, it no longer
tries to print out the section of src where it thinks the error is.
For comparison, running perl on this code file:
# this is line 1
my $x :foo(bar;
the
quick
brown
fox jumps over the lazy dog
used to output:
Unterminated attribute parameter in attribute list at /tmp/p1 line 0, near "\0\0x jumps "
syntax error at /tmp/p1 line 0, at EOF
Execution of /tmp/p1 aborted due to compilation errors.
but now outputs:
Unterminated attribute parameter in attribute list at /tmp/p1 line 2.
Note how previously: the error message included two literal null chars
(represented by \0 above), followed by a random chunk of the last line;
it claimed to be on line 0; it output two further error messages.
For comparison, change the ':foo' to 'q' so that its an unterminated
string, and you get (and always got):
Can't find string terminator ")" anywhere before EOF at /tmp/p1 line 2.
Sullivan Beck [Thu, 1 Sep 2016 15:08:40 +0000 (11:08 -0400)]
Bump Locale-Codes from 3.39 to 3.40
Steve Hay [Wed, 7 Sep 2016 07:49:54 +0000 (08:49 +0100)]
Upgrade XSLoader from version 0.22 to 0.24
Steve Hay [Wed, 7 Sep 2016 07:43:53 +0000 (08:43 +0100)]
Upgrade Sys::Syslog from version 0.34 to 0.35
This includes changes that supersede the blead customization.
Father Chrysostomos [Wed, 7 Sep 2016 05:11:05 +0000 (22:11 -0700)]
[perl #129106] Check for null PL_curcop in IN_LC()
or, rather, in macros that it calls.
When exiting a string eval, the current cop may be freed, so PL_curcop
gets set to null. With the -DC option, we may end up printfing NVs
during scope exit, so the locale macros used to make sure that the
locale is sane before converting the numbers to strings need to make
sure not to read PL_curcop->cop_hints when PL_curcop is null.
This used to crash with: ./miniperl -DC -e'eval "l/A"'
I’m not sure how to write a test for this, or even whether it’s worth
writing a test for -D, which often changes behaviour depending on
porters’ whims.
Tony Cook [Wed, 7 Sep 2016 01:16:49 +0000 (11:16 +1000)]
fix a stupid type error
Tony Cook [Mon, 22 Aug 2016 03:56:26 +0000 (13:56 +1000)]
(perl #128988) preserve PL_oldoldbufptr is preserved in scan_heredoc()
In some cases this is used in building error messages.
Tony Cook [Mon, 15 Aug 2016 00:39:22 +0000 (10:39 +1000)]
(perl #128095) check pack_sockaddr_un()'s return value
pack_sockaddr_un() silently truncates the supplied path if it won't
fit into the sun_path member of sockaddr_un.
This may change in the future, but for now check the path in the
sockaddr matches the desired path, and skip if it doesn't.
Father Chrysostomos [Mon, 5 Sep 2016 17:14:29 +0000 (10:14 -0700)]
Fix up B::Concise tests following op flag change
Father Chrysostomos [Mon, 5 Sep 2016 16:31:31 +0000 (09:31 -0700)]
[perl #47047] Fix erroneous AUTOLOAD warning
If there was a stub present in the package into which the invocant had
been blessed, then AUTOLOADing via a *method* call would warn with ‘Use
of inherited AUTOLOAD for non-method’ even if it is a method.
A recent commit stopped OPf_REF from being set on OP_ENTERSUB, so this
commit uses that flag to indicate a method call, to allow a fast run-
time check to see whether to pass the method flag to gv_autoload.
Rick Delaney [Mon, 5 Sep 2016 16:25:59 +0000 (09:25 -0700)]
Test for perl #47047
Father Chrysostomos [Mon, 5 Sep 2016 06:27:42 +0000 (23:27 -0700)]
No need to skip t/op/dump.t on darwin
Father Chrysostomos [Mon, 5 Sep 2016 06:14:21 +0000 (23:14 -0700)]
Stop setting OPf_REF on OP_ENTERSUB
It isn’t doing anything really here, and I need it for
another purpose.
David Mitchell [Mon, 5 Sep 2016 14:49:28 +0000 (15:49 +0100)]
toke.c: fix mswin32 builds
9bde56224 added this as part of macro:
- PL_last_lop_op = f; \
+ PL_last_lop_op = f < 0 ? -f : f; \
which broke win32 builds due to this
UNIBRACK(-OP_ENTEREVAL)
expanding to
PL_last_lop_op = -345 < 0 ? --345 : -345
and the -- being seen as a pre-dec op.
Diagnosed by Dagfinn Ilmari Mannsåker.
Craig A. Berry [Mon, 5 Sep 2016 02:13:28 +0000 (21:13 -0500)]
Make the "where install" directories Unix-format on VMS.
These paths will get converted to Unix format for loading into
@INC anyway, but since
483efd0abe3 they really need to start out
that way. Otherwise, when running under a Unix shell, the path
delimiter will be ':' and the absolute VMS specs will get split
in half in S_incpush_use_sep(), which kicks in before the
conversion to Unix format.
Craig A. Berry [Mon, 5 Sep 2016 02:10:06 +0000 (21:10 -0500)]
Tainted dirs on VMS when not under DCL.
Since
483efd0abe3 the path delimiter is a ':' instead of '|' on
VMS when running under a Unix shell. So use that as a guide
to whether we should use a colon or a slash to detect relative
directories that should be tainted.
David Mitchell [Mon, 5 Sep 2016 11:50:37 +0000 (12:50 +0100)]
newATTRSUB_x(): document what the cv var is for
David Mitchell [Mon, 5 Sep 2016 11:46:02 +0000 (12:46 +0100)]
make S_already_defined() in op.c return void
Currently it returns a boolean value, but after the removal of the MAD
code, this value is just equivalent to its 'block' arg; so instead just
make its callers test 'block' instead. This makes the calling code in
newATTRSUB_x and newMYSUB easier to follow.
Also, put the block condition on a single line (whitespace change)
David Mitchell [Mon, 5 Sep 2016 11:37:55 +0000 (12:37 +0100)]
reindent and reformat newMYSUB and newATTRSUB_x
This commit only contains whitespace changes.
As a follow-on from the previous commit, reindent a couple of block of
code; but also do a general re-indenting and partial refomatting of these
two similar functions, which were a bit of mess and hard to follow:
* replace weird mixtures of indents with standard 4-indents;
* split if (X) Y into two lines;
* separate out blocks with a blank line;
* outdent labels by exactly 2 spaces
David Mitchell [Mon, 5 Sep 2016 11:08:56 +0000 (12:08 +0100)]
newMYSUB/Perl_newATTRSUB_x remove a goto
In both these functions there is code like:
if (!block)
goto attrs;
...
attrs:
Change them to
if (block) {
...
}
attrs:
Lets not use gotos if there's a better way.
Will re-indent in next commit.
David Mitchell [Mon, 5 Sep 2016 10:52:23 +0000 (11:52 +0100)]
avoid using freed ops on BEGIN :attr {}
If a BEGIN sub has a code attribute applied (no idea why you would want to
do such a thing, but it's not illegal) then part of applying the attribute
is to do 'use attributes', which compiles
BEGIN { require "attributes"; attributes->import(AAA) }
so we end up compiling a BEGIN while in the middle of compiling a BEGIN.
The part of Perl_newATTRSUB_x() that under some circumstances copies
the body of the newly-compiled CV to the old CV which occupies the name
slot, kicks in here.
Since the ops that make up the AAA above were allocated from the old
BEGIN's op slabs, they get prematurely freed when the old BEGIN's
ops are discarded by the SvREFCNT_dec(PL_compcv).
The simplest fix is to just avoid the copy if we're compiling a BEGIN.
David Mitchell [Mon, 5 Sep 2016 10:15:49 +0000 (11:15 +0100)]
do_sv_dump(): handle CvSTART() as slab address
If a CV is CvSLABBED(), then CvSTART() points to the op slab rather than a
start op. Make Perl_do_sv_dump() display this more informatively.
David Mitchell [Sat, 27 Aug 2016 13:28:24 +0000 (14:28 +0100)]
assert op not freed in finalize_op() and rpeep()
This should never happen.
Father Chrysostomos [Mon, 5 Sep 2016 05:11:00 +0000 (22:11 -0700)]
parser.t: Suppress warning
Dan Collins [Sun, 4 Sep 2016 18:43:41 +0000 (14:43 -0400)]
Regression test for RT #129196
Father Chrysostomos [Mon, 5 Sep 2016 03:24:19 +0000 (20:24 -0700)]
[perl #129196] Crash/bad read with ‘evalbytes S’
5dc13276 added some code to toke.c that did not take into account
that the opnum (‘f’) argument to UNI* could be a negated op number.
PL_last_lop_op must never be negative, since it is used as an offset
into a struct.
Tests for the crash will come in the next commit.
Karl Williamson [Sat, 3 Sep 2016 18:17:17 +0000 (12:17 -0600)]
inline.h: Declare functions to be inline
The branch merged with commit
8d19ebbca9eecf219cc453cffe88722722860dd9
forgot to give this hint to the compiler.
Father Chrysostomos [Sun, 4 Sep 2016 23:02:02 +0000 (16:02 -0700)]
release_schedule.pod: Tick off last month