| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This was brought up in ticket #113812.
Formats that are nested inside closures only work if invoked from
directly inside that closure. Calling the format from an inner sub
call won’t work.
Commit af41786fe57 stopped it from crashing, making it work as well
as 5.8, in that closed-over variables would be undefined, being
unavailable.
This commit adds a variation of the find_runcv function that can check
whether CvROOT matches an argument passed in. So we look not for the
current sub, but for the topmost sub on the call stack that is a clone
of the closure prototype that the format’s CvOUTSIDE field points to.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
A closure prototype’s CvOUTSIDE pointer might have been modified if
its containing sub is freed first. When a sub is cloned, the enclos-
ing sub is always the currently-running sub (not so for formats).
So this commit makes subs always use find_runcv, the way they did
before 71f882da828.
So the closure logic which was needed for formats is now moved into an
else branch that is used only for them.
|
| |\
| |
| |
| |
| |
| |
| | |
This branch uses per-CV slabs for ops, so that ops can all be freed
after compilation errors, fixing memory leaks and a crash.
See commit 8be227ab5e for how it works.
|
| | | |
|
| | | |
|
| | | |
|
| | |
| |
| |
| |
| | |
If an op slot is reused for a smaller op, we only need to zero out the
space used for the op, not the whole slot.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
I was using Perl_warn, both for its convenience, and because the line
numbers were extremely helpful in tracking bugs.
But it invokes the warnhook, if present, and also respects tied
STDERR. We should be using Perl_debug_log.
Changing this also avoids the need for /* diag_listed_as: SKIPME */
all over the place.
|
| | | |
|
| | |
| |
| |
| | |
They require attributes.pm.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
For simplicity, op slots are never resized once they are allocated.
But they are reused after they are freed, if they are big enough.
When allocating the last op slot that will fit on a slab, we might as
well enlarge the slot to contain whatever space is left over, so this
slot, after being freed, can be reused for a larger op.
|
| | |
| |
| |
| |
| | |
Instead of using #ifndef every time we call cv_forget_slab, just
define it as a no-op under PL_OP_SLAB_ALLOC.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
I am having difficulty getting these tests to fail. They crash when
run standalone, but always pass when run via fresh_perl.t. Hopefully,
they will fail somewhere. :-)
Yes, fresh_perl.t does begin with this:
# ** DO NOT ADD ANY MORE TESTS HERE **
But t/comp/parser.t does not (and should not) use test.pl, so it is
very hard to test something like this.
Putting it here seemed slightly better than putting it in
its own file.
|
| | | |
|
| | | |
|
| | | |
|
| | |
| |
| |
| |
| |
| | |
These proved extremely useful for getting this slab allocator to work.
We might as well leave them in place for future debugging.
|
| | | |
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This addresses bugs #111462 and #112312 and part of #107000.
When a longjmp occurs during lexing, parsing or compilation, any ops
in C autos that are not referenced anywhere are leaked.
This commit introduces op slabs that are attached to the currently-
compiling CV. New ops are allocated on the slab. When an error
occurs and the CV is freed, any ops remaining are freed.
This is based on Nick Ing-Simmons’ old experimental op slab implemen-
tation, but it had to be rewritten to work this way.
The old slab allocator has a pointer before each op that points to a
reference count stored at the beginning of the slab. Freed ops are
never reused. When the last op on a slab is freed, the slab itself is
freed. When a slab fills up, a new one is created.
To allow iteration through the slab to free everything, I had to have
two pointers; one points to the next item (op slot); the other points
to the slab, for accessing the reference count. Ops come in different
sizes, so adding sizeof(OP) to a pointer won’t work.
The old slab allocator puts the ops at the end of the slab first, the
idea being that the leaves are allocated first, so the order will be
cache-friendly as a result. I have preserved that order for a dif-
ferent reason: We don’t need to store the size of the slab (slabs
vary in size; see below) if we can simply follow pointers to find
the last op.
I tried eliminating reference counts altogether, by having all ops
implicitly attached to PL_compcv when allocated and freed when the CV
is freed. That also allowed op_free to skip FreeOp altogether, free-
ing ops faster. But that doesn’t work in those cases where ops need
to survive beyond their CVs; e.g., re-evals.
The CV also has to have a reference count on the slab. Sometimes the
first op created is immediately freed. If the reference count of
the slab reaches 0, then it will be freed with the CV still point-
ing to it.
CVs use the new CVf_SLABBED flag to indicate that the CV has a refer-
ence count on the slab. When this flag is set, the slab is accessible
via CvSTART when CvROOT is not set, or by subtracting two pointers
(2*sizeof(I32 *)) from CvROOT when it is set. I decided to sneak the
slab into CvSTART during compilation, because enlarging the xpvcv
struct by another pointer would make all CVs larger, even though this
patch only benefits few (programs using string eval).
When the CVf_SLABBED flag is set, the CV takes responsibility for
freeing the slab. If CvROOT is not set when the CV is freed or
undeffed, it is assumed that a compilation error has occurred, so the
op slab is traversed and all the ops are freed.
Under normal circumstances, the CV forgets about its slab (decrement-
ing the reference count) when the root is attached. So the slab ref-
erence counting that happens when ops are freed takes care of free-
ing the slab. In some cases, the CV is told to forget about the slab
(cv_forget_slab) precisely so that the ops can survive after the CV is
done away with.
Forgetting the slab when the root is attached is not strictly neces-
sary, but avoids potential problems with CvROOT being written over.
There is code all over the place, both in core and on CPAN, that does
things with CvROOT, so forgetting the slab makes things more robust
and avoids potential problems.
Since the CV takes ownership of its slab when flagged, that flag is
never copied when a CV is cloned, as one CV could free a slab that
another CV still points to, since forced freeing of ops ignores the
reference count (but asserts that it looks right).
To avoid slab fragmentation, freed ops are marked as freed and
attached to the slab’s freed chain (an idea stolen from DBM::Deep).
Those freed ops are reused when possible. I did consider not reusing
freed ops, but realised that would result in significantly higher mem-
ory using for programs with large ‘if (DEBUG) {...}’ blocks.
SAVEFREEOP was slightly problematic. Sometimes it can cause an op to
be freed after its CV. If the CV has forcibly freed the ops on its
slab and the slab itself, then we will be fiddling with a freed slab.
Making SAVEFREEOP a no-op won’t help, as sometimes an op can be
savefreed when there is no compilation error, so the op would never
be freed. It holds a reference count on the slab, so the whole
slab would leak. So SAVEFREEOP now sets a special flag on the op
(->op_savefree). The forced freeing of ops after a compilation error
won’t free any ops thus marked.
Since many pieces of code create tiny subroutines consisting of only
a few ops, and since a huge slab would be quite a bit of baggage for
those to carry around, the first slab is always very small. To avoid
allocating too many slabs for a single CV, each subsequent slab is
twice the size of the previous.
Smartmatch expects to be able to allocate an op at run time, run it,
and then throw it away. For that to work the op is simply mallocked
when PL_compcv has’t been set up. So all slab-allocated ops are
marked as such (->op_slabbed), to distinguish them from mallocked ops.
All of this is kept under lock and key via #ifdef PERL_CORE, as it
should be completely transparent. If it isn’t transparent, I would
consider that a bug.
I have left the old slab allocator (PL_OP_SLAB_ALLOC) in place, as
it is used by PERL_DEBUG_READONLY_OPS, which I am not about to
rewrite. :-)
Concerning the change from A to X for slab allocation functions:
Many times in the past, A has been used for functions that were
not intended to be public but were used for public macros. Since
PL_OP_SLAB_ALLOC is rarely used, it didn’t make sense for Perl_Slab_*
to be API functions, since they were rarely actually available. To
avoid propagating this mistake further, they are now X.
|
| | |
| |
| |
| |
| |
| |
| |
| | |
This is a dummy op type that should never be seen by any code except
op allocation code (to come).
So it is not in the usual list of opcodes, but is #defined outside the
range valid of opcodes.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
This is to allow future commits to free dangling ops after errors.
If an op is on the savestack, then it is going to be freed by scope.c,
and op_free must not be called on it by anyone else.
So we flag such ops new.
|
| | | |
|
| |/
|
|
|
| |
This will indicate that a CV has a reference count on, and ownership
of, a slab used for allocating ops.
|
| |
|
|
|
|
|
| |
The default case for non-overloaded classes is fallback => 1, so saying
fallback => 1 on its own shouldn't enable overloading, but saying
fallback => undef on its own should (even though undef is the default
for overloaded classes).
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
This makes
package Foo;
use overload fallback => 0;
and
package Bar;
use overload '+' => \&add, fallback => 0;
behave identically when an operator other than '+' is used.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If a format closing over lexical variables is defined inside a clo-
sure, it must only be called directly inside that closure, not from
any other eval, sub, or format.
Calling it from anywhere else started causing a crash in 5.10.0,
because the format would try to close over variables in the currently-
running sub, using padoffsets intended for a completely unrelated pad.
This commit stops it from crashing by checking whether the currently-
running sub is a clone of the format’s outer sub (a closure proto-
type). If it is not, the outer closure prototype is used, resulting
in ‘Variable is not available’ warnings.
This makes things work as well as they did in 5.8. Ideally, we should
search the call stack for the topmost clone of the format’s outer sub;
but I’m saving that for another commit.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This began crashing in 5.11.3:
sub foo {
sub bar {
my ($a,$b,$c,$d,$e,$f,$g,$h,$i,$j,$k,$l,$m,$n,$o,$p,$q,$r,$s,$x);
format =
@||||||
$x
.
}
}
undef *bar;
write;
(On some systems, you need more alphabet soup to make it crash.)
This commit (just the perly.y part shown) caused it to crash:
commit 421f30ed1e95009450bdc7905bf3433ee806ea4f
Author: Zefram <zefram@fysh.org>
Date: Tue Dec 15 11:48:31 2009 +0100
[perl #22977] Bug in format/write
diff --git a/perly.y b/perly.y
index 18e5875..a61a6b3 100644
--- a/perly.y
+++ b/perly.y
@@ -511,7 +511,9 @@ peg : PEG
;
format : FORMAT startformsub formname block
- { SvREFCNT_inc_simple_void(PL_compcv);
+ {
+ CV *fmtcv = PL_compcv;
+ SvREFCNT_inc_simple_void(PL_compcv);
#ifdef MAD
$$ = newFORM($2, $3, $4);
prepend_madprops($1->tk_mad, $$, 'F');
@@ -521,6 +523,10 @@ format : FORMAT startformsub formname block
newFORM($2, $3, $4);
$$ = (OP*)NULL;
#endif
+ if (CvOUTSIDE(fmtcv) && !CvUNIQUE(CvOUTSIDE(fmtcv))) {
+ SvREFCNT_inc_simple_void(fmtcv);
+ pad_add_anon((SV*)fmtcv, OP_NULL);
+ }
}
;
Unfortunately, adding the format to the pad like that (to allow
pad_fixup_inner_anons to fix up formats as well as subs) is proble-
matic. It causes the format’s CvOUTSIDE to be weak. Since the for-
mat does not hold a reference count on its outer sub, that sub can be
freed before the format. When that happens, regular subs are fixed
up by having CvOUTSIDE change to point to the grandparent. If you
do that for formats, you run into a problem: Formats can be cloned
even when the outer sub is not running. Formats are cloned whenever
invoked *by name* via write. If CvOUTSIDE points to a different sub,
then closing over the scalars in specific pad offsets in that sub can
result in reading past the end of the pad. If you don’t read past the
end of the pad, you are still making variables close over unrelated variables, so the inner $x could close over an outer @y, etc. Subrou-
tines don’t have that problem, as they can only be cloned when they
have an outer sub. (Even though the outer sub’s prototype, if it is a
closure, might have been freed, the outer sub itself is still running
and referenced by the context stack.)
This commit changes the direction of the weak reference between an
outer sub’s pad and an inner format, fixing the crash.
To do so, it has to store, not the format itself, but a weak RV point-
ing to the format, in the outer sub’s pad.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
amagic_call now does its best to propagate the operator's context into
the overload callback. It's not always possible - for instance,
dereferencing and stringify/boolify/numify always have to return a
value, even if it's not used, due to the way the overload callback works
in those cases - but the majority of cases should now work. In
particular, overloading <> to handle list context properly is now
possible.
For backcompat reasons (amagic_call and friends are technically public
api functions), list context will not be propagated unless specifically
requested via the AMGf_want_list flag. If this is passed, and the
operator is called in list context, amagic_call returns an AV* holding
all of the returned values instead of an SV*. Void context always
results in amagic_call returning &PL_sv_undef.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
| |
bison-2.5.1 adds less superfluous semicolons at the end of action blocks,
but works fine.
|
| |
|
|
| |
Also, prevent a crash when SvCUR is used on a non-PV.
|
| | |
|
| |
|
|
|
|
| |
./perl -Ilib t/op/sort.t won't do what you want, because tests that
don't use Test::More don't set an error code on their own, you need to
run it under the harness
|
| |\
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Quoting Brian Fraser:
So.
I more or less went ahead and did this, to some extent. To some other
extent the changes are stylistic, like consistently using v5.x.y instead of
5.x or 5.x.y or whatnot in perlvar (and only in perlvar, so as to avoid
bikeshedding). I also bumped most odd versions, post 5.6, to the next
stable release (so, for example, 5.7.1 became 5.8.0, and 5.9.0 became
5.10.0, but 5.005 is still 5.005).
There's only one big deletion, which was most of perltrap.pod -- It dealt
with traps for Perl 4 programmers migrating to Perl 5. It would be really
swell if someone could update perltrap for their other favorite language of
choice, since right now the document is severely lacking, only dealing with
awk, shell, C/C++, and Perl itself.
There's some stuff that I didn't really know about, so I left it alone,
including all of perlhack(tips)? and perl(re|deb)?guts.
One thing that came up in IRC while I was doing this is that having a table
in perlvar detailing in which version each variable became available would
be really swell. And I know that brian d foy compiled one for string/regex
escapes, which strikes me like a good candidate to get in as well.
|
| | |
| |
| |
| | |
Regened known_pod_issues.dat for the previous commits.
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
