| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
| |
This new script tests that goto &xsub causes the sub to see the hints,
not of the subroutine it replaces, but of that subroutine’s caller.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
commit 3c78429c102e0fe2ad30c60dfe52636b6071ef19 reduced the depth
of recursion in rpeep(), by deferring recursion into branches until a bit
later (so that the recursive call to rpeep was then likely to be shallow).
However, it went one step further: when the chain of op_next's had been
exhausted in the main loop, it processed any remaining deferrred branches
in the main loop rather than recursing. All nice and efficient, but it
broke the expectation that someone who had hooked into rpeep could follow
the chain of op_nexts in each call and visit *all* ops.
This commit removes that optimisation and restores the rpeep hook
expectancy. This shouldn't have any major effect on the depth of
recursion, and its minor inefficiency doesn't really matter for a
one-time compilation-time pass.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Long blocks of code that include logical or loop ops (i.e. those with
multiple 'branches' of ops, such as op_other, op_redo etc) cause
Perl_rpeep to recurse deeply and eventaully SEGV.
For example this crashes, due to the ENTERLOOP:
eval ("{\$x = 1 }\n" x 10000)
The deep recursion happens because the processing of the entire rest of
the code occurs in within the nested call. For example in the code
A && B; C; D; E;
the ops are structured as
A -> AND -> C -> D -> E
\ /
B
where AND->op_next points to C, while AND->op_other points to B.
rpeep() would normally process each op in the op_next sequence in turn
(i.e. A/AND/C/D/E), but when it reaches AND, it recursively calls
rpeep(B), which happens to then process B/C/D/E. Finally it returns,
and the parent rpeep processes C, finds it's already done, and exits.
Clearly, if C,D,E etc also contain conditional/loop ops, then the
recursion level gradually stacks up.
The fix for this is to add a small deferred queue to rpeep().
Whenever rpeep wants to recurse with op_other or op_lastop etc,
it instead adds it to the deferred queue. Only if the queue is full is
rpeep actually called. The hope is that by deferring, when we do
eventually process it, enough of the main op_next chain has already been
processed to ensure that the child rpeep returns very early.
In the example above, processing of AND causes B to be added to the queue,
and the main rpeep process continues processing C, D etc. Sometime later,
the queue becomes full and B is processed via a recursive call to rpeep.
B is processed, and op_next is followed to C, but C is marked as already
processed, so the child rpeep returns almost immediately.
For LOOP ops, I've stopped following op_redoop and op_nextop, since
AFAIKT the ops these point to will also be reachable vie op_next anyway.
op_lastop is the exception; in while(1){..} only op_lastop points to the
rest of the code block.
Note that this commit doesn't guarantee only shallow recursion, it just
makes deep recursion fairly unlikely.
Note also that this commit causes the order of the processing of op_next
chains to be altered; this can affect the ordering of compiler warnings
and fatal messages among potentially other things.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
SvIsCOW was ignoring the fact that it might be passed a
typeglob, which made its behaviour contradict its docs.
This fixes that and, in doing so, simplifies the
upcoming Internals::SvREADONLY fix.
|
| |
|
|
|
| |
CLONEf_COPY_STACKS is only used by the windows pseudo-fork.
This test allows testing/debugging of CLONEf_COPY_STACK without needing threads or Windows.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This fixes a regression introduced since 5.14.0, by commit e0171a1a3.
The new Perl_hfree_next_entry function that that commit introduced
returns the value of the hash element, or NULL if there are none left.
If the value of the hash element is NULL, the two cases are indistin-
guishable.
Before e0171a1a3, all the hash code took null values into account.
mro_package_moved took advantage of that, stealing values out of a
hash and leaving it to the freeing code to delete the elements.
The two places that call Perl_hfree_next_entry (there was only one,
S_hfreeentries, with commit e0171a1a3, but the following commit,
104d7b699c, made sv_clear call it, too) were not accounting for NULL
values’ being returned, and could terminate early, resulting in mem-
ory leaks.
One could argue that the perl core should not be assigning nulls to
HeVAL, but HeVAL is part of the public API and there could be CPAN
code assigning NULL to it, too.
So the safest approach seems to be to modify Perl_hfree_next_entry’s
callers to check the number of keys and not to attribute a signifi-
cance to a returned NULL.
|
| | |
|
| | |
|
| |
|
|
|
| |
including the ‘Useless assignment to a temporary’ warning
which is only triggered by these.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Before this commit:
commit f07ec6dd59215a56bc1159449a9631be7a02a94d
Author: Zefram <zefram@fysh.org>
Date: Wed Oct 13 19:05:19 2010 +0100
remove filter inheritance option from lex_start
The only uses of lex_start that had the new_filter parameter false,
to make the new lexer context share source filters with the previous
lexer context, were uses with rsfp null, which therefore never invoked
source filters. Inheriting source filters from a logically unrelated
file seems like a silly idea anyway.
string evals could inherit the same source filter space as the cur-
rently compiling code. Despite what the quoted commit message says,
sharing source filters allows filters to be inherited in both direc-
tions: A source filter created when the eval is being compiled also
applies to the file with which it is sharing its space.
There are at least 20 CPAN distributions relying on this behaviour
(or, rather, what could be considered a Test::More bug). So this com-
mit restores the source-filter-sharing capability. It does not change
the current API or make public the API for sharing source filters, as
this is supposed to be a temporary stop-gap measure for 5.14.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
| |
There are four TODO tests in caller.t, relating to the hints hash returned
under the debugger. Since 20439bc77dfeec46d94a15cf108446039e26c995, two
of these tests has started to pass. I don't understand this area enough
to know whether that commit *should* have made them pass, but it
seems fairly consistent. So un-TODO just those two.
This means that under the debugger, we now expect to get a hints hash, but
that its contents are still wrong.
|
| |
|
|
|
|
|
|
|
| |
# New Ticket Created by (Peter J. Acklam)
# Please include the string: [perl #81882]
# in the subject line of all future correspondence about this issue.
# <URL: http://rt.perl.org/rt3/Ticket/Display.html?id=81882 >
Signed-off-by: Abigail <abigail@abigail.be>
|
| |
|
|
|
|
| |
New API functions parse_fullexpr(), parse_listexpr(), parse_termexpr(),
and parse_arithexpr(), to parse an expression at various precedence
levels.
|
| | |
|
| | |
|
| |
|
|
| |
correctly
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Convert sv_eq_flags() and sv_cmp_flags() to use it.
Previously, to compare two strings of characters, where was was in UTF-8, and
one was not, you had to either:
1: Upgrade the second to UTF-8
2: Compare the resulting octet sequence
3: Free the temporary UTF-8 string
or:
1: Attempt to downgrade the first to bytes. If it can't be, they aren't equal
2: Else compare the resulting octet sequence
3: Free the temporary byte string
Which for the general case involves a malloc()/free() and at least two O(n)
scans per comparison.
Whereas this approach has no allocation, a single O(n) scan, which terminates
as early as the best case for the second approach.
|
| |
|
|
|
|
|
|
|
| |
Attached patch fixes the following warning from 'make test':
ext/XS-APItest/t/overload...................................."my"
variable $got masks earlier declaration in same scope at t/overload.t
line 81.
ok
|
| |
|
|
| |
tryAMAGICunDEREF() isn't used anywhere in the core. Add tests for it.
|
| |
|
|
|
|
|
|
| |
Commit 2f8edad0d37e91319b6ba10b3745327ea49c179 made G_ARRAY
equal to G_SCALAR | G_VOID, contrary to perlcall.pod.
Bring the documentation up to date and add a test to prevent a similar
(although unlikely) accident in the future.
|
| |
|
|
|
|
|
|
|
| |
This removes around 300 bytes of object code from each place it was previously
inlined. It also provides a better interface - quite a lot of the core
currently bodges things by creating a local variable C<SV **sp = &sv> to use
the macro.
Change the XS::APItest wrapper to amagic_deref_call().
|
| | |
|
| |
|
|
|
|
| |
The function scope() goes into the API as op_scope(), and mod() goes
into the API as op_lvalue(). Both marked experimental, because their
behaviour is a little quirky and not trivially dequirkable.
|
| |
|
|
|
|
|
| |
New API function parse_label() parses a label, separate from statements.
If a label has not already been lexed and queued up, it does not use
yylex(), but parses the label itself at the character level, to avoid
unwanted lexing past an absent optional label.
|
| |
|
|
|
| |
New API function parse_barestmt() parses a pure statement, with no label,
and returns just the statement's core ops, not attaching a state op.
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
When $1 had the utf8 flag set from a previous match, SvPVbyte
may croak with 'Wide character in subroutine entry' before
resetting the flag to its new value.
Add a support function and a TODO test for this in XS-APItest.
http://bugs.debian.org/376329
|
| |
|
|
|
|
|
|
| |
Dying and returning from a format both relied on the state of a
just-popped context frame being preserved across a LEAVE. Don't rely
on it. Test using an operator ripped off from Scope::Cleanup, which makes
it easy to run arbitrary Perl code during cleanup, without isolating it
on a separate context stack as the DESTROY mechanism does.
|
| |
|
|
|
|
|
|
|
|
|
| |
'make test' produces the following warning:
ext/XS-APItest/t/multicall.....................................Useless
use of private variable in void context at t/multicall.t line 37.
ok
The attached patch fixes this by adding a "no warnings 'void';"
statement to the test file.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Expose cop hint hashes as a type COPHH, with a cophh_* API which is a
macro layer over the refcounted_he_* API. The documentation for cophh_*
describes purely API-visible behaviour, whereas the refcounted_he_*
documentation describes the functions mainly in terms of the
implementation. Revise the cop_hints_* API, using the flags parameter
consistently and reimplementing in terms of cophh_*. Use the cophh_*
and cop_hints_* functions consistently where appropriate.
[Modified by the committer to update two calls to
Perl_refcounted_he_fetch recently added to newPMOP.]
|
| |
|
|
|
| |
New API function parse_block() parses a code block, including surrounding
braces. The block is a lexical scope, but not inherently a dynamic scope.
|
| |
|
|
|
| |
Set PL_expect at the start of parse_fullstmt() as well as at the start
of parse_stmtseq(). Test both.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
When recursing into the parser for recursive-descent parsing, put
a special entry on the bracket stack that generates a fake EOF if a
closing bracket belonging to an outer parser frame is seen. This keeps
the bracket stack balanced across a parse_stmtseq() frame, fixing
[perl #78222].
If a recursive-descent parser frame ends by yyunlex()ing an opening
bracket, pop its entry off the bracket stack and stash it in the
forced-token queue, to be revived when the token is re-lexed. This keeps
the bracket stack balanced across a parse_fullstmt() frame.
|
| |
|
|
|
|
|
|
| |
Syntax plugins can modify the source being parsed. It's fine for
them to modify the lexer buffer, but this must not be the same scalar
that was supplied to lex_start() and may be in use outside. Therefore
always copy the scalar in lex_start() rather than just referencing it.
Fixes [perl #78358].
|
| |
|
|
|
|
|
|
| |
See [perl #78070].
Basically, POPSUB/LEAVESUB had a mechanism to decrement the reference
count of the CV only at CvDEPTH==1; POP_MULTICALL was decrementing it at
all depths.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The macros dMULTICALL, PUSH_MULTICALL, MULTICALL and POP_MULTICALL
are completely untested in core apart from incidentally in List-Util.
The exercise they get there is probably quite comprehensive, but it's
not explicitly testing the macros themselves.
Add a hook and new test file to XS::APItest specifically for this purpose.
Currently the test file is almost empty.
The multicall_each function is shamelessly stolen from List:;Util::first.
|
| |
|
|
|
| |
Also rename the underlying function to op_linklist, to match the other
API op functions.
|
| |
|
|
|
|
| |
Put into the API op_append_elem, op_prepend_elem, and op_append_list. All
renamed from op_-less internal names. Parameter types for op_append_list
changed to match the rest of the op API and avoid some casting.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
New magic type PERL_MAGIC_checkcall attaches a function to a CV, which
will be called as the second half of the op checker for an entersub
op calling that CV. Default state, in the absence of this magic,
is to process the CV's prototype if it has one, or apply list context
to all the arguments if not. New API functions cv_get_call_checker()
and cv_set_call_checker() provide a clean interface to this facility,
hiding the internal use of magic.
Expose in the API the new functions rv2cv_op_cv(),
ck_entersub_args_list(), ck_entersub_args_proto(), and
ck_entersub_args_proto_or_list(), which are meaningful segments of
standard entersub op checking and are likely to be useful in plugged-in
call checker functions.
Expose new API function op_contextualize(), which is a public interface
to the internal scalar()/list()/scalarvoid() functions. This API is
likely to be required in most plugged-in call checker functions.
Incidentally add new function mg_free_type(), in the API, which will
remove magic of one type from an SV. (mg_free() removes all magic,
and there isn't anything else more selective.)
|
| | |
|
| |
|
|
|
|
|
| |
It's unlikely that XS_VERSION will contain a bogus version string (for long),
but the value passed in (or derived from $XS_VERSION or $VERSION) might well.
For that case, without this change, temporary SVs created within
xs_version_bootcheck() won't be freed (before interpreter exit).
|
| | |
|
