Father Chrysostomos [Sat, 28 Jul 2012 07:33:33 +0000 (00:33 -0700)]
Oust sv_gmagical_2iv_please
The magic flags patch prevents this from ever being called, since the
OK flags work the same way for magic variables now as they have for
muggle vars, avoid these fiddly games. (It was when writing it that I
realised the value of the magic flags proposal.)
Nicholas Clark [Fri, 27 Jul 2012 17:17:54 +0000 (19:17 +0200)]
Remove code for supporting 80286 based systems.
The 80286 was released two years before Perl 1, but the support code was
added with Perl 3. The chip hasn't been produced for more than 15 years -
even the 80386 hasn't been manufactured since 2007. Most of the other
memory model code was removed by commit
5869b1f143426909 in Sep 2000, so
support for 16 bit systems is long dead.
Nicholas Clark [Fri, 27 Jul 2012 15:53:22 +0000 (17:53 +0200)]
Remove dead code related to the Atari ST port of perl 4.0 patchlevel 19
The subdirectory containing the port specific files was purged when 5.000
was released, but changes made to other files were not removed.
Nicholas Clark [Sat, 28 Jul 2012 08:11:08 +0000 (10:11 +0200)]
Merge the refactoring of the filetest OPs' return code.
Nicholas Clark [Fri, 27 Jul 2012 14:38:01 +0000 (16:38 +0200)]
Comment the code with how filetest operators interact with the Perl stack.
Nicholas Clark [Fri, 27 Jul 2012 13:05:00 +0000 (15:05 +0200)]
Eliminate the macros FT_RETURN_FALSE() and FT_RETURN_TRUE().
As they now simply return the results of S_ft_return_false() and
S_ft_return_true() respectively, use this explicitly in the 7 places where
the macros had been used.
Nicholas Clark [Fri, 27 Jul 2012 12:47:56 +0000 (14:47 +0200)]
Refactor the macro FT_RETURN_TRUE() into the function S_ft_return_true()
This makes the true and false code paths as similar as possible. It also
eliminates a macro that the HP compiler was choking on.
Nicholas Clark [Fri, 27 Jul 2012 12:19:27 +0000 (14:19 +0200)]
Consolidate the code for returning false values from filetest operators.
Move the code that implements the non-stacking return processing from the
macro FT_RETURN_FALSE() into the static function S_ft_stacking_return_false().
Rename the function to S_ft_return_false() as it now handles all false cases.
Nicholas Clark [Fri, 27 Jul 2012 11:32:33 +0000 (13:32 +0200)]
Reorder S_ft_stacking_return_false().
Move the code which deals with setting the return value on perl's stack
ahead of the code which calculates which op to run next (the return value of
the C function).
Nicholas Clark [Fri, 27 Jul 2012 11:16:29 +0000 (13:16 +0200)]
Remove dSP from all filetest ops.
Following commit
d2c4d2d1e22d3125, all filetest ops avoid manipulating the
stack pointer until the point of pushing a return value. As all the filetest
returns now go through either FT_RETURN_FALSE() or FT_RETURN_TRUE(), it's
viable to put the dSP inside those two macros and use *PL_stack_sp in place
of TOPs. This eliminates all uses of sp in the bodies of the functions
(explicit, or implicit via macros), and that makes it clear that the main
bodies of the functions are not manipulating the stack.
Nicholas Clark [Fri, 27 Jul 2012 10:54:49 +0000 (12:54 +0200)]
Replace the macro RETURNX() with its expansion in FT_RETURN_{FALSE,TRUE}
The macros FT_RETURN_FALSE and FT_RETURN_TRUE in pp_ctl.c are already very
complex, sufficient to trigger an internal failure in HP's compiler [and
possibly also some humans :-)]. Replacing RETURNX() with the 3 statements it
expands to makes the intent of macros clearer, and exposes more refactoring
possibilities.
Father Chrysostomos [Sat, 28 Jul 2012 07:17:04 +0000 (00:17 -0700)]
perldelta for
1eb0b7be2ff1 (B::Deparse and loopex prec)
Father Chrysostomos [Sat, 28 Jul 2012 07:15:58 +0000 (00:15 -0700)]
perldelta for
1f039d60d3 (last $foo)
Father Chrysostomos [Sat, 28 Jul 2012 07:12:59 +0000 (00:12 -0700)]
perldelta for
2ba1f20ac3a (loopex prec docs)
Father Chrysostomos [Sat, 28 Jul 2012 07:10:53 +0000 (00:10 -0700)]
Father Chrysostomos [Sat, 28 Jul 2012 07:09:11 +0000 (00:09 -0700)]
perldelta for
42409c4069 (truncate FILENAME)
Father Chrysostomos [Sat, 28 Jul 2012 07:07:17 +0000 (00:07 -0700)]
Father Chrysostomos [Sat, 28 Jul 2012 07:04:25 +0000 (00:04 -0700)]
perldelta for
c9df4fdaad9 (dump LABEL leak)
Father Chrysostomos [Sat, 28 Jul 2012 06:59:19 +0000 (23:59 -0700)]
doop.c: Simplify do_trans’ un-cow logic
Since it calls SvPV_force_nomg a little further on, there is no need
for a separate sv_force_normal call to handle COWs.
Father Chrysostomos [Sat, 28 Jul 2012 06:51:59 +0000 (23:51 -0700)]
Fix C++ build broken by
1f039d60d3
The goto was bypassing initialisation. Avoiding goto altogether actu-
ally simplifies things.
Father Chrysostomos [Sat, 28 Jul 2012 06:46:07 +0000 (23:46 -0700)]
Flatten vstrings modified in place
A substitution forces its target to a string upon successful substitu-
tion, even if the substitution did nothing:
$ ./perl -Ilib -le '$a = *f; $a =~ s/f/f/; print ref \$a'
SCALAR
Notice that $a is no longer a glob after s///.
But vstrings are different:
$ ./perl -Ilib -le '$a = v102; $a =~ s/f/f/; print ref \$a'
VSTRING
I fixed this in 5.16 (
1e6bda93) for those cases where the vstring ends
up with a value that doesn’t correspond to the actual string:
$ ./perl -Ilib -le '$a = v102; $a =~ s/f/o/; print ref \$a'
SCALAR
It works through vstring set-magic, that does the check and removes
the magic if it doesn’t match.
I did it that way because I couldn’t think of any other way to fix
bug #29070, and I didn’t realise at the time that I hadn’t fixed
all the bugs.
By making SvTHINKFIRST true on a vstring, we force it through
sv_force_normal before any in-place string operations. We can also
make sv_force_normal handle vstrings as well. This fixes all the lin-
gering-vstring-magic bugs in just two lines, making the vstring set-
magic (which is also slow) redundant. It also allows the special case
in sv_setsv_flags to be removed.
Or at least that was what I had hoped.
It turns out that pp_subst, twists and turns in tortuous ways, and
needs special treatment for things like this.
And do_trans functions wasn’t checking SvTHINKFIRST when arguably it
should have.
I tweaked sv_2pv{utf8,byte} to avoid copying magic variables that do
not need copying.
Father Chrysostomos [Sat, 28 Jul 2012 01:05:02 +0000 (18:05 -0700)]
pp.c:pp_trans: avoid redundant sv in transr
I think I added the if (OP_TRANSR) block to the wrong spot, because the
mortal scalar created just before it is only used for y/// without /r.
Eric Brine [Fri, 27 Jul 2012 22:54:24 +0000 (18:54 -0400)]
Document return to be exempt from the looks-like-a-function rule, like dump, goto and next.
Father Chrysostomos [Fri, 27 Jul 2012 21:52:21 +0000 (14:52 -0700)]
perlfunc: document last/next EXPR
Also remove some repetitive text from goto, added in
2ba1f20ac3.
Father Chrysostomos [Fri, 27 Jul 2012 21:41:05 +0000 (14:41 -0700)]
B::Deparse: loopexes have assignment prec
See ticket #113684 for detail.
Father Chrysostomos [Fri, 27 Jul 2012 21:37:22 +0000 (14:37 -0700)]
Increase $B::Deparse::VERSION to 1.16
Father Chrysostomos [Fri, 27 Jul 2012 21:26:39 +0000 (14:26 -0700)]
[perl #113684] Make redo/last/next/dump accept expr
These functions have been allowing arbitrary expressions, but would
treat anything that did not resolve to a const op as the empty string.
Not only were arguments swallowed up without warning, but constant
folding could change the behaviour. Computed labels are allowed for
goto, and there is no reason to disallow them for these other ops.
This can also come in handy for certain types of code generators.
In the process of modifying pp functions to accept arbitrary labels,
I noticed that the label and loop-popping code was identical in three
functions, so I moved it out into a separate static function, to make
the changes easier.
I also had to reorder newLOOPEX significantly, because code under the
goto branch needed to a apply to last, and vice versa. Using multiple
gotos to switch between the branches created too much of a mess.
I also eliminated the use of SP from pp_last, to avoid copying the
value back and forth between SP and PL_stack_sp.
Father Chrysostomos [Fri, 27 Jul 2012 20:36:35 +0000 (13:36 -0700)]
pp.h: Make [TP]OPp and [TP]OPpx identical
In the absence of n_a (see
8c074e2a and
95fad918), there is no differ-
ence between [TP]OPp and [TP]OPpx except speed, so there is no reason
for the x-less variant to be deprecated.
Father Chrysostomos [Fri, 27 Jul 2012 20:02:13 +0000 (13:02 -0700)]
op.c: apidoc typo
Father Chrysostomos [Fri, 27 Jul 2012 17:46:26 +0000 (10:46 -0700)]
[perl #113684] Document actual prec of loop exits
These have always* had assignment precedence, such that
$a = goto $b = $c
is equivalent to
$a = (goto ($b = $c))
* I haven’t checked before perl 5.
Father Chrysostomos [Fri, 27 Jul 2012 17:30:58 +0000 (10:30 -0700)]
Fix CvOUTSIDE assert/refcnt bugs with sub redefinition
my $sub = sub { 4 };
*foo = $sub;
*bar = *foo;
undef &$sub;
eval "sub bar { 3 }";
undef *foo;
undef *bar;
As of 5.8.4, this script produces:
Attempt to free unreferenced scalar: SV 0x8002c4.
As of 5.14.0:
panic: del_backref.
Or, undef debugging builds:
Assertion failed: (!CvWEAKOUTSIDE(cv)), function Perl_newATTRSUB_flags, file op.c, line 7045.
Commit
5c41a5fa918 (backported to 5.8.4 in commit
7a565e5d) caused the
first bug:
commit
5c41a5fa918d32924e1ac2f02418d5d7f465ef26
Author: Dave Mitchell <davem@fdisolutions.com>
Date: Sun Jan 25 02:04:23 2004 +0000
Remove small memory leak in newATTRSUB that manifested as a
leaking scalar after the interpeter was cloned
p4raw-id: //depot/perl@22209
diff --git a/op.c b/op.c
index b902fed..5fd21bf 100644
--- a/op.c
+++ b/op.c
@@ -4165,6 +4165,8 @@ Perl_newATTRSUB(pTHX_ I32 floor, OP *o, OP *proto, OP *attrs, OP *block)
/* transfer PL_compcv to cv */
cv_undef(cv);
CvFLAGS(cv) = CvFLAGS(PL_compcv);
+ if (!CvWEAKOUTSIDE(cv))
+ SvREFCNT_dec(CvOUTSIDE(cv));
CvOUTSIDE(cv) = CvOUTSIDE(PL_compcv);
CvOUTSIDE_SEQ(cv) = CvOUTSIDE_SEQ(PL_compcv);
CvOUTSIDE(PL_compcv) = 0;
Checking the flags right after clobbering them can’t be a good idea.
Commit
437388a93 caused the panics and assertion failures. See com-
mit f6894bc for detail.
Commit f6894bc fixed the panics and assertion failures involving CvGV.
One remaining assertion (!CvWEAKOUTSIDE) added by
437388a93 is still
incorrect. It’s not true that CvWEAKOUTSIDE is never set on a re-
used stub.
In both cases (
5c41a5fa’s code and
437388a93’s code), the weakness
of CvOUTSIDE is ignored and the outside sub (the eval) is freed
prematurely.
It could be that this type of redefinition should be disallowed (des-
pite its usefulness), but that is a separate issue. This used to
work. And pure-Perl code should not be triggering assertion failures
or freeing scalars twice.
Father Chrysostomos [Fri, 27 Jul 2012 17:11:25 +0000 (10:11 -0700)]
Fix CvGV assertion bug with sub redefinition
*foo = \&baz;
*bar = *foo;
eval 'sub bar { print +(caller 0)[3], "\n" }';
bar();
Before 5.14, that produces:
main::foo
As of 5.14, it produces:
main::baz
Or, under debugging builds:
Assertion failed: (CvGV(cv) == gv), function Perl_newATTRSUB_flags, file op.c, line 7139.
commit
437388a93114b1acbfb3a173dfa7aa2138fd8283
Author: Nicholas Clark <nick@ccl4.org>
Date: Thu Nov 18 14:54:44 2010 +0000
Refactor newATTRSUB()'s logic for grafting a sub definition to an existing stub
Previously it was using cv_undef() to (partially) free the target CV (the
pre-existing stub), before donating it the padlist and outside pointers from
the source CV (the definition, just compiled), and then freeing up the remains
of the source CV.
Instead, explicitly exchange padlist and outside pointers, explicitly assign
other fields that need changing (file and stash), and assert that various
CvFLAGS are as we expect them.
That commit adds some assertions, including:
+ assert(!CvCVGV_RC(cv));
+ assert(CvGV(cv) == gv);
Those assertions are not always true. CvGV might be refcounted, and
it might not point to the same gv.
437388a93 also changed things such that the CVf_CVGV_RC flag is clob-
bered, so refcounting and backrefs get out of synch (tests for that
specific bug will be in a subsequent commit). It also stopped sub
redefinition from setting CvGV.
Nicholas Clark [Fri, 27 Jul 2012 15:36:40 +0000 (17:36 +0200)]
Restore microperl, which has been unable to build since January.
microperl was broken by commit
82ad65bb0613be64 on 2012-01-16, which used
IN_LOCALE_COMPILETIME for the first time in the C code. Unlike
IN_LOCALE_RUNTIME, it had no fallback definition for microperl.
Commit
f90a9a0230170cc0 on 2012-01-28 added the first use of Strtol(),
which means that microperl now needs to assume that the system has strtol().
(Which is not unreasonable, as it's part of C89).
Chip Salzenberg [Thu, 26 Jul 2012 03:27:30 +0000 (20:27 -0700)]
When setting environment variables via %ENV, force values to be strings only
(turning off other OK flags), make them byte strings; if wide characters can't
be downgraded to bytes, leave the string utf8 and issue a warning.
Chip Salzenberg [Thu, 26 Jul 2012 01:41:50 +0000 (18:41 -0700)]
Make all the SvPV*force* macros always return sv with SvPOK_only, as API
docs always claimed they did. Also update those docs to be clearer.
Karl Williamson [Thu, 26 Jul 2012 15:56:27 +0000 (09:56 -0600)]
mktables: Change \w definition to match new Unicode's
Unicode has changed their definition of what should match \w.
http://www.unicode.org/reports/tr18/. This follows that change.
Chris 'BinGOs' Williams [Thu, 26 Jul 2012 14:58:02 +0000 (15:58 +0100)]
Make Module::CoreList install into 'site' >= 5.012
Also if versiononly is set make sure that corelist is installed
with the appropriate versioned suffix.
Father Chrysostomos [Thu, 26 Jul 2012 07:16:00 +0000 (00:16 -0700)]
[perl #113872] Fix leavewrite’s stack handling
This commit fixes Scope::Escape compatibility by restoring the old
stack pointer stored on the context stack when exiting a write.
I don’t really understand why this fixes Scope::Escape, or rather, how
Scope::Escape ends up leaving some number of items on the stack other
than 1. But I *do* know this is the correct approach, as it mirrors
what pp_leavesub does in scalar context, but pops the stack back
to the old value (SP = newsp), rather than the old value+1 (see
pp_hot.c:pp_leavesub: MARK = newsp + 1; and later SP = MARK;). Then
the code that follows takes care of pushing write’s own return value.
Father Chrysostomos [Thu, 26 Jul 2012 05:19:56 +0000 (22:19 -0700)]
op.c: op_clear is tempting fate
This if() statement can be reached by op types to which the OpTRANS*
flags to not apply. They happen at present not to use any flags that
conflict with these (except when OPf_KIDS is set, in which case this
code is not reached). But we should make sure, via an assertion,
that new flags added to goto or last do not conflict with trans utf8
flags, and that trans utf8 flags (1 and 2), if renumbered, do not con-
flict with goto/last utf8 flags (128).
Father Chrysostomos [Thu, 26 Jul 2012 05:07:12 +0000 (22:07 -0700)]
Don’t let ?: folding affect truncate
truncate(${\1} ? foo : bar, 0) and truncate(1 ? foo : bar, 0) should
behave the same way, but were treated differently, due to the way ?:
is folded in the latter case. Now that foldedness is recorded in the
op tree (
cc2ebcd7902), we can use the OPpCONST_FOLDED flag to distin-
guish truncate(1 ? foo : bar, 0) from truncate(foo, 0).
Father Chrysostomos [Thu, 26 Jul 2012 05:04:02 +0000 (22:04 -0700)]
Stop truncate(word) from falling back to file name
In commit
5e0adc2d66, which was a bug fix, I made the mistake of
checking the truth of the return value of gv_fetchsv, which is called
when truncate’s argument is a bareword.
This meant that truncate FOO, 0; would truncate the file named FOO if
the glob happened to have been deleted.
Father Chrysostomos [Thu, 26 Jul 2012 03:15:36 +0000 (20:15 -0700)]
Don’t let ?: folding affect stat
stat(${\1} ? foo : bar) and stat(1 ? foo : bar) should behave the same
way, but were treated differently, due to the way ?: is folded in
the latter case. Now that foldedness is recorded in the op tree
(
cc2ebcd7902), we can use the OPpCONST_FOLDED flag to distinguish
stat(1 ? foo : bar) from stat(foo).
Father Chrysostomos [Thu, 26 Jul 2012 01:00:34 +0000 (18:00 -0700)]
Merge ck_trunc and ck_chdir
ck_chdir, added in 2006 (
d4ac975e) duplicates ck_trunc, added in
1993 (
79072805), except for a null op check which is harmless when
applied to chdir.
Father Chrysostomos [Wed, 25 Jul 2012 23:40:37 +0000 (16:40 -0700)]
op.c: dump LABEL leaks its label
./perl -Ilib -e 'warn $$; eval "sub { dump a }" while 1'
Watch the memory usage go up.
It didn’t have its own case in op_clear.
Father Chrysostomos [Wed, 25 Jul 2012 23:31:07 +0000 (16:31 -0700)]
op.c:op_free: Rmv dead code; simplify cop_free logic
This reverts
c53f1caa and
cc93af5f.
See the thread starting at
http://www.nntp.perl.org/group/perl.perl5.porters/2008/04/msg135885.html
Basically, change
c53f1caa made a change, but then
cc93af5f undid it,
but differently. This resulted in dead code left by
c53f1caa (type is
unused after the assignment). And in the end the code behaved exactly
the same way, so the original problem was not fixed. I suspect this
was a B::C bug.
Karl Williamson [Wed, 25 Jul 2012 03:39:32 +0000 (21:39 -0600)]
Merge branch 'blead' of ssh://perl5.git.perl.org/perl into blead
Karl Williamson [Wed, 25 Jul 2012 03:14:46 +0000 (21:14 -0600)]
Merge branch 'khw/invlist' into blead
This topic branch deals mostly with bracketed character classes in
regular expressions. It has several main thrusts:
1) The character class macros in handy.h are tied to the class numbers
in regcomp.h, and a new set of regnode types, POSIX, are introduced.
This will allow more table driven code in regular expression
compilation and matching, so that the same regnode type can be used
for any of the Posix-like character classes, such as \w as well as
[:upper:]. This will allow removal of nearly-duplicate (and
triplicate, etc) code.
2) The optimizations for bracketed character classes are extended to
work off not just the first 256 characters, but all code points.
This extends some char class optimizations to Unicode, and will
allow future work to change the regular expression optimizer to work
off all Unicode characters, instead of its current behavior of mostly
giving up on these.
3) Several new character class optimizations are introduced, with the
groundwork laid for more. A character class containing a single
character is the same as the single character without the class,
except that most metacharacters are treated as literals. Thus,
/a[.]b/ produces identical results to /a\.b/. Prior to this merge,
the first form's compiled version would require quite a bit more
space than the second. Now they are identical.
Inversion lists of the whole class are used for optimization
calculations.
Karl Williamson [Mon, 23 Jul 2012 04:37:21 +0000 (22:37 -0600)]
regcomp.c: Revise bracketed char class optimizations
This commit uses the inversion list instead of the bit map constructed
during compilation of [bracketed classes] for determining the
optimizations that are done at the very end of processing the class.
This provides optimizations for things that can't be seen by just
looking at the bitmap.
There are optimizations done earlier in the code for things
that can be easily caught in Pass 1, but now we have complete
information. At this time, I'm not repeating checking for most
optimizations that are checked for earlier, though this could be added.
The earlier optimizations could overlook cases where someone specified a
class in a suboptimal way. For example the earlier code looks for a
class with a single range [0-9], but those 10 code points could instead
have been specified via [
0123456789], and the code doesn't catch that
currently; here we could, but I'm not doing so at this time.
The code does do some duplicate checking. For example, some Unicode
properties match only a single code point, such as lb=cr, and these
aren't known at the earlier point where single code point classes are
checked for.
Karl Williamson [Mon, 23 Jul 2012 04:09:21 +0000 (22:09 -0600)]
regcomp.c: Fix <if> condition
The else clause is expecting that the regex is compiled under /d, when
in fact, until this commit, it could also be under /l. I could not come
up with a case currently where this distinction matters, but it's best
to not tempt fate.
Karl Williamson [Mon, 23 Jul 2012 04:00:46 +0000 (22:00 -0600)]
regcomp.c: Add _invlist_contains_cp
This simply searches an inversion list without going through a swash.
It will be used in a future commit.
Karl Williamson [Wed, 25 Jul 2012 02:53:41 +0000 (20:53 -0600)]
utf8.c: Add a get_() method to hide internal details
This should have been written this way to begin with (I'm the culprit).
But we should have a method so another routine doesn't have to know the
internal details.
Karl Williamson [Fri, 20 Jul 2012 17:39:09 +0000 (11:39 -0600)]
regcomp.c: Optimize /[[:blank:]]/u into \h
These two are equivalent.
Karl Williamson [Fri, 20 Jul 2012 17:35:07 +0000 (11:35 -0600)]
regcomp.c: Properly count elements in [] for false ranges
When something that looks like a range turns out not to be, the hyphen
is matched literally, and so is a separate element in the character
class. It needs to be accounted as such.
Karl Williamson [Fri, 20 Jul 2012 16:23:14 +0000 (10:23 -0600)]
regcomp.c: Use POSIXA, NPOSIXA
This commit optimizes character classes which are matched under /a or
/aa and consist of a single Posix class, into POSIXA or NPOSIXA regop
types. For example /[[:word:]]/a. Since [:ascii:] is always
ascii-restricted no matter what the charset modifier is, it is always
optimized.
These nodes should execute somewhat faster than a generic ANYOF node,
and are significantly smaller, taking 2 bytes instead of 12.
The flags field of the node structure is used to hold an enum indicating
which of the 15 Posix classes is being matched.
Karl Williamson [Fri, 20 Jul 2012 16:02:23 +0000 (10:02 -0600)]
regcomp.sym: Add new node types POSIXA and NPOSIXA
These will be used to handle things like /[[:word:]]/a. This patch
doesn't add the code to actually use these. That will be done in a
future patch.
Also, placeholders POSIXD, POSIXL, and POSIXU are also added for future
use.
Karl Williamson [Mon, 9 Jul 2012 21:17:28 +0000 (15:17 -0600)]
regcomp.h: Use handy.h constants
This synchronizes the ANYOF_FOO usages to the isFOO() usages. Future
commits will take advantage of this relationship.
Karl Williamson [Fri, 20 Jul 2012 03:53:06 +0000 (21:53 -0600)]
handy.h: Free up bits in PL_charclass[]
This array is a bit map containing the Posix and similar character
classes for the first 256 code points. Prior to this commit many
character classes were represented by two bits, one for characters that
are in it over the full Latin-1 range, and one for just the ASCII
characters that are in it. The number of bits in use was approaching
the 32-bit limit available without playing games.
This commit takes advantage of a recent commit that adds a bit to the
table for all the ASCII characters, and the fact that the ASCII
characters in a character class are a subset of the full Latin1
range. So, iff both the full-range character class bit and the ASCII
bit is set is that character an ASCII-range character with the given
character class.
A new internal macro is created to generate code to determine if a
character is an ASCII range character with the given class. It's not
clear if the generated code is faster or slower than the full range
version.
The result is that nearly half the bits are freed up, as the ones for
the ASCII-range are now redundant.
Karl Williamson [Fri, 20 Jul 2012 03:47:48 +0000 (21:47 -0600)]
handy.h: Add intermediate internal macro
This macro abstracts an operation, and will make future commits cleaner.
Karl Williamson [Tue, 10 Jul 2012 04:00:16 +0000 (22:00 -0600)]
regcomp.c: Relax some restrictions on optimizations for locale
Prior to this commit, we didn't do any inversions for bracketed
character classes running under locale. However, this is more strict
than necessary. If there is no folding, and everything else is known at
compile time, then what is matched when the result is complemented is
well-defined, and can be done now. (Also clarifies one of the affected
comments)
Karl Williamson [Tue, 10 Jul 2012 00:44:46 +0000 (18:44 -0600)]
regcomp.c: Add func to test 2 inversion lists for equality
This adds _invlistEQ which for now is commented out
Karl Williamson [Mon, 9 Jul 2012 22:35:11 +0000 (16:35 -0600)]
utf8.c: Add info to commented-out -DU lines
This proved useful when I recently needed to use these for debugging
Karl Williamson [Mon, 9 Jul 2012 22:17:40 +0000 (16:17 -0600)]
regcomp.c: Reverse order of setting, for speed
It's faster to append to an inversion list than to insert into the
middle. The previous order of doing things guaranteed that the 2nd
thing done would be an insertion, hence slower than an append. Now we
add the lowest ordinal character first, so there is a chance that both
will be appends
Karl Williamson [Mon, 9 Jul 2012 21:12:36 +0000 (15:12 -0600)]
perllocale: Mention that \n doesn't change for locales
Karl Williamson [Mon, 9 Jul 2012 20:30:29 +0000 (14:30 -0600)]
handy.h: Remove duplicated test
This test is duplicated in the called macro
Karl Williamson [Mon, 9 Jul 2012 20:28:11 +0000 (14:28 -0600)]
handy.h: White space only
This moves a #define next to similar ones, and removes some white space
Karl Williamson [Mon, 9 Jul 2012 20:25:03 +0000 (14:25 -0600)]
regcomp.c: Move table to wider scope
This table will be used in future commits outside this block's scope
Karl Williamson [Mon, 9 Jul 2012 19:59:09 +0000 (13:59 -0600)]
regcomp.c: Silence compiler warning
I suspect this being an IV stemmed from an earlier version. It always
contains unsigneds
Karl Williamson [Mon, 9 Jul 2012 19:54:13 +0000 (13:54 -0600)]
regcomp.c: Change macro name to better indicate its purpose
Karl Williamson [Sat, 7 Jul 2012 16:10:30 +0000 (10:10 -0600)]
Optimize a single character [class] into EXACTish
Things like /[s]/ or /[s]/i can be optimized as if they did not have the
brackets, /s/ or /s/i.
Karl Williamson [Mon, 9 Jul 2012 19:10:00 +0000 (13:10 -0600)]
regcomp.c: Extract some code into an inline function
This code will be used in future commits in multiple places
Karl Williamson [Sat, 7 Jul 2012 17:46:40 +0000 (11:46 -0600)]
regcomp.c: shrink some optimized [class] nodes
Various bracketed character class specifications don't need the full
generality, and can be optimized into smaller, faster nodes. Recent
commits, such as
3a64b5154fffec75126d34d25954f0aef30d9f8a have
introduced some such optimizations. However, their commit messages are
wrong, in that they don't end taking up any less space than the original
ANYOF node, as there is no provision for giving it back, once reserved.
This commit corrects that for non-locale nodes. Restructuring of the
code, more than I care to do now, would be required to extend this to
locale nodes.
Only optimizations that are determined in pass1 of the regex compilation
are eligible for this, as once the space is calculated, it is reserved
before pass2 begins. There are now two sections where optimization is
done in this routine. The final section is after all the data is
examined and processed, and we know exactly what is to be in the ANYOF
node. Currently, most of this calculation processing is skipped in pass
1. We could do everything in both passes, greatly slowing down the
first, in order to figure out the exact space requirements.
But instead, I've chosen to add a separate optimization section that
looks at only the optimizations that are easily computable in the
current first pass, and to apply those early, in time for the shrinking
to occur. This commit adds that shrinking.
Locale nodes can't be shrunk because the code currently writes into
their larger buffer before this decision to optimize is made. To
illustrate, say that the node is at the end of the regex, and the data
is written at x+20, within a generic locale ANYOF node's space. Pass 1
doesn't write anything, but calculates the space needed. At the point
that this commit addresses, we would shrink the amount of space to be
allocated to be allocated. Then we carve out space for the regex from
the heap. That x+20 could be pointing to something entirely different,
which Pass 2 would destroy. This can be avoided by storing the stuff
that would be written into a temporary until the optimization decision
has been made. But there is extra work involved in doing this, that I
don't think is worth it at this time.
The same problem doesn't occur in non-locale situations, even though the
flags field of the node is written before the optimization decision.
This is because every node has the space for a flags field, so it just
gets overwritten as necessary when the optimization is done.
Karl Williamson [Sat, 7 Jul 2012 17:42:30 +0000 (11:42 -0600)]
regcomp.c: Delay some initialization until needed
This delays the initialization of the bitmap in ANYOF nodes until just
before it is needed, and to after where we make a decision to optimize
that node to a node which takes less space. Currently, the space is
not given up, once reserved in pass 1, so the write is harmless. This
will allow a future commit to shrink the space.
Karl Williamson [Sat, 7 Jul 2012 14:15:56 +0000 (08:15 -0600)]
regcomp.c: Remove duplicate assignments
These variables are already set to the same values a few lines up.
Karl Williamson [Tue, 3 Jul 2012 04:18:47 +0000 (22:18 -0600)]
handy.h: Move bit shifting into base macro
This changes the #defines to be just the shift number, while doing
the shifting in the macro that the number is passed to. This will prove
useful in future commits
Karl Williamson [Tue, 3 Jul 2012 03:39:56 +0000 (21:39 -0600)]
handy.h: Renumber character class bits
These are renumbered so that the ones that correspond to character
classes in regcomp.h are related numerically as well. This will prove
useful in future commits.
Karl Williamson [Tue, 3 Jul 2012 03:31:31 +0000 (21:31 -0600)]
handy.h: Reorder some #defines
They are now ordered in the same order as the similar #defines in
regcomp.h. This will be useful in later commits
Karl Williamson [Tue, 3 Jul 2012 03:09:18 +0000 (21:09 -0600)]
handy.h: l1_charclass.h: Add bit for matching ASCII
This does not replace the isASCII macro definition, as I think the
current one is more efficient than this one provides. But future
commits will rely on all the named character classes (e.g.,
/[[:ascii:]]/) having a bit, and this is the only one missing.
Karl Williamson [Tue, 3 Jul 2012 03:04:16 +0000 (21:04 -0600)]
handy.h: refactor some macros to use a new one in common.
This creates a new, unpublished, macro to implement most of the other
macros. This macro will be useful in future commits.
Karl Williamson [Sat, 7 Jul 2012 03:41:07 +0000 (21:41 -0600)]
regcomp.c: Extract code to inline function
Future commits will use this paradigm in additional places, so extract
it to a function, so they all do things right. This isn't a great API,
but it works for the few places this will be called.
Karl Williamson [Fri, 6 Jul 2012 21:14:19 +0000 (15:14 -0600)]
regcomp.sym: Correct and add comments
Karl Williamson [Fri, 6 Jul 2012 17:31:11 +0000 (11:31 -0600)]
regen/regcomp.pl: Allow ';' in comments
If a comment contained a semi-colon, the regular expression's greedy
quantifier would think the portion of the comment before it was part of
the data to be processed
Karl Williamson [Wed, 4 Jul 2012 21:22:59 +0000 (15:22 -0600)]
regcomp.c: Optimize [^\n] into \N
This optimization is a big win, as it takes less space, and completely
avoids any swash hash creation, and use.
Karl Williamson [Wed, 4 Jul 2012 20:49:22 +0000 (14:49 -0600)]
regcomp.c: White-space, comments only
This fixes some nits in a some comments, adds some comments, but mostly
just indents and outdents to reflect a new outer block, and to fit
within 80 columns
Karl Williamson [Wed, 4 Jul 2012 20:16:29 +0000 (14:16 -0600)]
regcomp.c: Refactor new charclass optimizations
Commits
3a64b5154fffec75126d34d25954f0aef30d9f8a and
3172e3fd885a9c54105d3b6156f18dc761fe29e5 introduced some optimizations
into the handling of bracketed character classes with just a single
element.
In working on other optimizations, I realized that it would be better to
put these all in one spot, instead of doing things partially and setting
flags to pass to other areas of the routine.
This commit moves all the processing to a single spot, which gets called
only after we know that there will be just one element in the character
class.
I also realized that the [0-9] optimization should strictly not be done
under locale. There is no test for this, as actually this would only be
a problem if a locale was in violation of the C standard. But (most) of
the rest of Perl doesn't assume that locales are well-behaved in this
regard, so now this code doesn't either.
Karl Williamson [Tue, 3 Jul 2012 02:50:33 +0000 (20:50 -0600)]
regcomp.c: Use data structure properties to remove tests
The ANYOF_foo character class #defines really form an enum, with the
property that the regular one is n, and its complement is n+1. So
we can replace the tests in each case: of the switch, with a single test
afterwards.
Karl Williamson [Sun, 1 Jul 2012 18:37:04 +0000 (12:37 -0600)]
regcomp.c: Use macro for clarity
It's kind of obscure what this test is doing, so change it into a macro
to clarify things
Karl Williamson [Thu, 28 Jun 2012 23:13:18 +0000 (17:13 -0600)]
regcomp.c: Only try to fold foldable chars
This uses the new table that a previous commit caused mktables to
generate which more precisely determines which Unicode characters
participate in folds. This avoids trying to calculate a fold that we
can know beforehand doesn't have a fold.
Karl Williamson [Thu, 28 Jun 2012 19:32:17 +0000 (13:32 -0600)]
mktables: Generate new table for foldable chars
This table consists of all characters that participate in any way in a
fold in the current Unicode version. regcomp.c currently uses the Cased
property as a proxy for these. This information is used to limit the
number of characters whose folds have to be dealt with in compiling
bracketed regex character classess. It turns out that Cased contains
more than 1300 more code points than actually do appear in folds, which
means potential extra work for compiling. Hence this patch allows that
work to be avoided.
There are a few characters in this new table that aren't in Cased, which
are potential bugs in the old way of doing things. In Unicode 6.1,
these are: U+02BC MODIFIER LETTER APOSTROPHE, U+0308 COMBINING
DIAERESIS, U+0313 COMBINING COMMA ABOVE, and U+0342 COMBINING GREEK
PERISPOMENI. I can't figure out how these might be currently causing a
bug, but this patch fixes any such.
Karl Williamson [Wed, 27 Jun 2012 23:18:20 +0000 (17:18 -0600)]
regcomp.c: Correct outdated comment
Karl Williamson [Wed, 27 Jun 2012 22:57:25 +0000 (16:57 -0600)]
regcomp.c: Change and comment a problematic #define
As noted in the new comments, there is no out-of-bounds code point, so
this should not be used as such; and it turns out, it really isn't.
But this is a fall-back in case code is changed to not set the variable
that this initializes, and the comment will prevent someone else from
wasting their time trying to figure out why this is like this.
Karl Williamson [Wed, 27 Jun 2012 22:37:56 +0000 (16:37 -0600)]
embed.fnc: Remove duplicate entry
Karl Williamson [Wed, 27 Jun 2012 22:31:36 +0000 (16:31 -0600)]
utf8.h, regcomp.c: Use mnemonics for Unicode chars
Add a mnemonic definition for these three code points. They are
currently used in only one place, but future commits will use them
elsewhere.
Karl Williamson [Wed, 25 Jul 2012 00:35:40 +0000 (18:35 -0600)]
embed.fnc: Add const to remove compiler warning
This should have been declared const.
Nicholas Clark [Tue, 24 Jul 2012 14:52:10 +0000 (16:52 +0200)]
In Perl_magic_setenv() s/ptr/key/ in two pieces of platform-specific code.
These were missed in commit
1203306491d341ed, which renamed ptr to key.
David Mitchell [Tue, 24 Jul 2012 14:11:29 +0000 (15:11 +0100)]
lex_grow_linestr(): update re_eval_start
As seen in [perl #114242], the TryCatch test suite triggers this bug.
lex_grow_linestr() was reallocing the PL_linestr buffer and updating all
pointers within it, but was failing to update
PL_sublex_info.re_eval_start.
Unfortunately, lex_grow_linestr() doesn't seem to be used in core outside
of hereto docs, and due to a separate parser bug, hereto docs can't
currently be used within /(?{})/, so I can't add tests for it yet.
Dave Rolsky [Fri, 20 Jul 2012 16:28:07 +0000 (11:28 -0500)]
Update core to Time::Local 1.2300 from CPAN
David Mitchell [Tue, 24 Jul 2012 10:21:25 +0000 (11:21 +0100)]
[perl #114248] Newxz? doc wrong in perlclib.pod
Commit
9f653bb54868b8547466924d4ce483acb8987efb changed the docs from
New/Newz to the newer Newx/Newxz, but forgot to remove the
no-longer-present 'id' arg.
Nicholas Clark [Tue, 24 Jul 2012 08:39:08 +0000 (10:39 +0200)]
Merge B 1.37 to blead (more SV flag constants exposed).
Nicholas Clark [Mon, 23 Jul 2012 11:39:05 +0000 (13:39 +0200)]
Bump B's VERSION and note the changes in perldelta.pod.
Nicholas Clark [Mon, 23 Jul 2012 11:27:32 +0000 (13:27 +0200)]
Expose all GV flags matching qr/GVf_/ as constants in B.
Previously only those matching qr/GVf_IMPORTED/ were exposed.