This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perl5.git
11 years ago[PATCH] Cleanup output from t/io/defout.t RT#78572
Jerry D. Hedden [Tue, 26 Oct 2010 21:11:32 +0000 (22:11 +0100)]
[PATCH] Cleanup output from t/io/defout.t RT#78572

  t/io/defout.t pollutes the output of 'make test':

  t/io/defout....................................................$a;
  # write
  ok

  The attached patch fixes this.

Signed-off-by: Chris 'BinGOs' Williams <chris@bingosnet.co.uk>
11 years agofix thinko in code comment
David Mitchell [Tue, 26 Oct 2010 16:07:57 +0000 (17:07 +0100)]
fix thinko in code comment

11 years agoUpdate CPAN to CPAN version 1.94_62
Chris 'BinGOs' Williams [Tue, 26 Oct 2010 13:41:47 +0000 (14:41 +0100)]
Update CPAN to CPAN version 1.94_62

  [DELTA]

  2010-10-26  Andreas J. Koenig  <andk@cpan.org>

        * release 1.94_62

        * address RT #62064: build_requires_install_policy set to "no" did not
        work correctly (reported by Xavier Bergade)

        * address RT ##55091: don't ask the proxy credentials if proxy_user
        empty (fixed by Robert Bohne)

        * address RT #55093: no_proxy doesn't work with more then one entries
        (fixed by Robert Bohne)

11 years agodetect the failure introduced in 304474c3 and fixed in cefd5c7c
Tony Cook [Tue, 26 Oct 2010 12:09:45 +0000 (23:09 +1100)]
detect the failure introduced in 304474c3 and fixed in cefd5c7c

11 years agoIn I18N::Langinfo::langinfo(), use newSVpv(), as it will call strlen()
Nicholas Clark [Tue, 26 Oct 2010 04:50:14 +0000 (05:50 +0100)]
In I18N::Langinfo::langinfo(), use newSVpv(), as it will call strlen()

newSVpv() will return an undef if the passed pointer is NULL, which allows
langinfo() to be simplified further.

11 years ago-$zero should not modify $zero
Father Chrysostomos [Tue, 26 Oct 2010 04:29:24 +0000 (21:29 -0700)]
-$zero should not modify $zero

With change a5b92898, negation started modifying numeric arguments,
causing problems for modules like Data::Float.

11 years agoCorrectly set param->new_perl in perl_clone_using()
Jan Dubois [Tue, 26 Oct 2010 01:35:19 +0000 (18:35 -0700)]
Correctly set param->new_perl in perl_clone_using()

11 years agoCorrect spelling
Father Chrysostomos [Tue, 26 Oct 2010 00:58:45 +0000 (17:58 -0700)]
Correct spelling

I misspelt lopadotemachoselachogaleokranioleipsanodrimypotrimmatosilphiokarabomelitokatakechymenokichlepikossyphophattoperisteralektryonoptokephalliokinklopeleiolagoiosiraiobaphetraganopterygon.

11 years agoRemove the to-do mark from test 4 in threads-dirh.t
Father Chrysostomos [Tue, 26 Oct 2010 00:56:48 +0000 (17:56 -0700)]
Remove the to-do mark from test 4 in threads-dirh.t

11 years agoadd CvSTASH_set() macro and make CvSTASH() rvalue only
Zefram [Mon, 25 Oct 2010 22:34:23 +0000 (23:34 +0100)]
add CvSTASH_set() macro and make CvSTASH() rvalue only

Now that CvSTASH requires backreference bookkeeping, stop people from
directly assigning to it (by using CvSTASH() as an lvalue), and instead
force them to use CvSTASH_set().

11 years agoDocument rewinddir() limitations
Jan Dubois [Mon, 25 Oct 2010 23:43:51 +0000 (16:43 -0700)]
Document rewinddir() limitations

as well as the similar limitations of directory handles that
have been open while calling the fork() emulation on Windows.

11 years agoFix @ISA recursion during global destruction
Father Chrysostomos [Mon, 25 Oct 2010 21:48:56 +0000 (14:48 -0700)]
Fix @ISA recursion during global destruction

With -DDEBUGGING and make test (not make test_harness):

Failed 3 tests out of 1889, 99.84% okay.
mro/basic.t
mro/recursion_c3.t
mro/recursion_dfs.t

See <http://www.nntp.perl.org/group/perl.perl5.porters/2010/10/msg165311.html>.

This was broken by change 35759254, which made S_hv_delete_common call
mro_package_moved when the G_DISCARD flag was passed. It used to be
skipped, which was a mistake.

This commit simply skips mro_package_moved when $::{"main::"} is
deleted. It was buggy anyway (main::E being passed to recursive calls,
instead of E), and should only happen at global destruction. (If
$::{"main::"} is deleted at run-time, it stop Perl from being func-
tional, so no one should be doing that.)

11 years agoImplement DIR* cloning on Windows
Jan Dubois [Mon, 25 Oct 2010 22:50:27 +0000 (15:50 -0700)]
Implement DIR* cloning on Windows

There doesn't seem to be a mechanism to clone FileFind handles on
Windows.  Therefore this implementation just reads all remaining
entries into a cache buffer and closes the handle.  All further
readdir() requests will be fulfilled from the cache buffer, in both
the original and the new interpreter.

This fixes bug 75154 on Windows (all tests in t/op/threads-dirh.t pass).

This commit also changes the return value of win32_telldir() to -1
for directory handles that have been read until the end.  The previous
return value was (NULL - dirp->start), which technically is not
valid C code.

API change alert: Perl_dirp_dup() gets an additional CLONE_PARAMS
parameter in this change (like all the other Perl_*_dup() functions).

11 years ago[perl #77812] Globs in tied scalars can be reified if STORE dies
Father Chrysostomos [Mon, 25 Oct 2010 21:21:57 +0000 (14:21 -0700)]
[perl #77812] Globs in tied scalars can be reified if STORE dies

This commit just adds tests. This was fixed by 2acc3314e31a9.

11 years agoRevert "Fix for RT#1804: Anonymous glob breaks when assigned through"
Father Chrysostomos [Mon, 25 Oct 2010 21:20:57 +0000 (14:20 -0700)]
Revert "Fix for RT#1804: Anonymous glob breaks when assigned through"

This reverts commit 0fe688f528b0e1b5bef6fb30d5e45316430e8a41, except
for the tests.

It is no longer necessary, as of change 2acc3314e31a9. Another reason
for this revert is that doing so fixes bug #77812, or at least it
would if 2acc3314e31a9 had not fixed that, too.

11 years agoMake untie check the FAKE flag on globs
Father Chrysostomos [Mon, 25 Oct 2010 19:28:33 +0000 (12:28 -0700)]
Make untie check the FAKE flag on globs

This allows untie($scalar) to untie the scalar if the last assigned or
returned happened to be a typeglob.

11 years agoIn Socket::inet_ntoa(), use Perl_newSVpvf(), instead of sprintf and newSVpvn*.
Nicholas Clark [Mon, 25 Oct 2010 19:43:57 +0000 (20:43 +0100)]
In Socket::inet_ntoa(), use Perl_newSVpvf(), instead of sprintf and newSVpvn*.

11 years agofunction to parse isolated label
Zefram [Sun, 24 Oct 2010 04:44:43 +0000 (05:44 +0100)]
function to parse isolated label

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.

11 years agofunction to parse unlabelled statement
Zefram [Sun, 24 Oct 2010 02:13:39 +0000 (03:13 +0100)]
function to parse unlabelled statement

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.

11 years agostop passing line numbers into op constructor functions
Zefram [Sat, 23 Oct 2010 14:40:58 +0000 (15:40 +0100)]
stop passing line numbers into op constructor functions

Remove the line number parameter from newWHILEOP() and newFOROP()
functions.  Instead, the line number for the impending COP is set by
parser code after constructing the ops.  (In fact the parser was doing
this anyway in most cases.)  This brings newWHILEOP() and newFOROP()
in line with the other op constructors, in that they do not concern
themselves with COPs.

11 years agorefactor and regularise label/statement grammar
Zefram [Tue, 19 Oct 2010 20:16:11 +0000 (21:16 +0100)]
refactor and regularise label/statement grammar

Refactoring of the grammar around statements.  New production <barestmt>
encompasses a statement without label.  It includes all statement types,
including declarations, with no unnecessary intermediate non-terminals.
It generates an op tree for the statement's content, with no leading
state op.  The <fullstmt> production has just one rule, consisting of
optional label followed by <barestmt>.  It puts a state op on the front
of the statement's content ops.

To support the regular statement op structure, the op sequence for for(;;)
loops no longer has a second state op between the initialisation and
the loop.  Instead, the unstack op type is slightly adapted to achieve
the stack clearing without a state op.

The newFOROP() constructor function no longer generates a state op,
that now being the job of the <fullstmt> production.  Consequently it
no longer takes a parameter stating what label is to go in the state op.
This brings it in line with the other op constructors.

11 years agoMerge Hash::Util::FieldHash::_test_uvar_{get,set,same} using ALIAS.
Nicholas Clark [Mon, 25 Oct 2010 19:21:50 +0000 (20:21 +0100)]
Merge Hash::Util::FieldHash::_test_uvar_{get,set,same} using ALIAS.

On this platform it reduces the object code size by over 1K.
Ideally these test functions would not be in the installed code, but I can't
see an easy way to avoid this.

11 years agoIn Hash::Util::FieldHash::_test_uvar_set, the second SvRV(sverf) is not needed.
Nicholas Clark [Mon, 25 Oct 2010 19:06:36 +0000 (20:06 +0100)]
In Hash::Util::FieldHash::_test_uvar_set, the second SvRV(sverf) is not needed.

11 years ago[perl #77688] tie $scalar can tie a handle
Father Chrysostomos [Mon, 25 Oct 2010 17:52:42 +0000 (10:52 -0700)]
[perl #77688] tie $scalar can tie a handle

Make tie() distinguish between $scalar and *$scalar when $scalar holds
a glob copy, by checking the FAKE flag.

11 years agoTests for [perl #77508] List assignment to fake globs fails
Father Chrysostomos [Mon, 25 Oct 2010 17:04:29 +0000 (10:04 -0700)]
Tests for [perl #77508] List assignment to fake globs fails

This was fixed by 2acc3314e31a9.

11 years ago[perl #77498] Assignment ignores magick when the RHS holds a glob
Father Chrysostomos [Mon, 25 Oct 2010 17:00:47 +0000 (10:00 -0700)]
[perl #77498] Assignment ignores magick when the RHS holds a glob

11 years agoFor PROXYSUBS, tables may sometimes have no entries.
Jan Dubois [Mon, 25 Oct 2010 17:23:04 +0000 (10:23 -0700)]
For PROXYSUBS, tables may sometimes have no entries.

The reason being that all entries in the tables are guarded
with #ifdef, so if none of the symbols are defined, then we
still end up with an empty table.  This actually happens in
the POSIX module on Windows for the "yes" table.

This reverts commit c553cca0 and changes based on it.

11 years agoIn FastCalc.xs, use SV * as the return type for _new() and _zero()
Nicholas Clark [Mon, 25 Oct 2010 13:53:19 +0000 (14:53 +0100)]
In FastCalc.xs, use SV * as the return type for _new() and _zero()

Previously these were using AV *. Whilst the code always was managing
references correctly, to work around the deficiency of the standard typemap
entry for AV * it had to mortalise the array it created, meaning that it was
creating 2 entries per call on the mortals stack, not 1.

11 years agort #72398 - get magic before downgrading in SvPVbyte()
Tony Cook [Mon, 25 Oct 2010 11:59:59 +0000 (22:59 +1100)]
rt #72398 - get magic before downgrading in SvPVbyte()

11 years agoTODO test: SvPVbyte should handle get magic before checking the utf8 flag
Niko Tyni [Mon, 25 Oct 2010 11:37:49 +0000 (22:37 +1100)]
TODO test: SvPVbyte should handle get magic before checking the utf8 flag

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

11 years agoConvert Math::BigInt::FastCalc to XSLoader.
Nicholas Clark [Mon, 25 Oct 2010 09:09:09 +0000 (10:09 +0100)]
Convert Math::BigInt::FastCalc to XSLoader.

As the minimum perl version is 5.006, enable 'use warnings'.

11 years agoIn FastCalc.xs, inline the macros CONSTANT_OBJ and RETURN_MORTAL_BOOL.
Nicholas Clark [Mon, 25 Oct 2010 08:53:44 +0000 (09:53 +0100)]
In FastCalc.xs, inline the macros  CONSTANT_OBJ and RETURN_MORTAL_BOOL.

With the previous commit, both are now only used in one place.
No need to use sv_2mortal() on the reset of boolSV(), as both PL_sv_no and
PL_sv_yes are immortals. [And special-cased within the implementation of
sv_2mortal() - not only is it a no-op, it's a non-free no-op :-)]

11 years agoReduce repetition in FastCalc.xs by using ALIASes.
Nicholas Clark [Mon, 25 Oct 2010 08:24:06 +0000 (09:24 +0100)]
Reduce repetition in FastCalc.xs by using ALIASes.

_one, _two and _ten are aliases to _zero
_is_odd is an alias of _is_even
_is_one, _is_two, _is_ten are aliases of _is_zero

On this system this reduces the object code size by about 4.5K (about 20%).

11 years agoMove some GV tests to gv.t from glob.t - the latter is for testing glob()
Nicholas Clark [Mon, 25 Oct 2010 07:54:07 +0000 (08:54 +0100)]
Move some GV tests to gv.t from glob.t - the latter is for testing glob()

They were incorrectly added to glob.t by 010be86bae505fbe.

11 years agoStop sysio.t changing into t/op, and hence creating a temporary file there.
Nicholas Clark [Mon, 25 Oct 2010 07:51:55 +0000 (08:51 +0100)]
Stop sysio.t changing into t/op, and hence creating a temporary file there.

Creating a file there causes a race condition with (at least) glob.t, which
lists the contents of t/op by various means, and compares the results for
consistency.

11 years agoCorrect tests for B::{cstring,perlstring,cchar}
Nicholas Clark [Mon, 25 Oct 2010 07:46:53 +0000 (08:46 +0100)]
Correct tests for B::{cstring,perlstring,cchar}

Fix a Perl precedence error that prevented several test cases for cchar() from
being run. As a result, fix those test cases.

Fix an ASCII assumption in the tests common to cstring() and perlstring().
(chr 127 is " on EBCDIC, which is most definitely printable, so won't be
replaced with octal there.)

11 years agoRename stashes when they move around
Father Chrysostomos [Mon, 25 Oct 2010 06:45:49 +0000 (23:45 -0700)]
Rename stashes when they move around

This is yet another patch in preparation for [perl #75176] (I keep
saying that.).

It uses the recently-added functions hv_name_add and hv_name_delete, to add and remove names when mro_package_moved is called.

mro_package_moved’s calling convention needed to change to make this
work, which is the bulk of the patch.

Code that was calling mro_package_moved was also doing it sometimes
when it was unnecessary. If the stash being assigned over had no name,
then there was no possibiiity of its being in the symbol table.

This probably fixes [perl #77358] (isa warnings), though I have not
tested that yet.

One user-visible change this introduces is that a detached glob whose
stash loses its name will no longer stringify the same way (a bit like
a glob that loses its stash pointer; except that it becomes
*__ANON__::foo instead of "").

11 years agoMemory-management macros evaluate their arguments multiple times
Father Chrysostomos [Mon, 25 Oct 2010 01:39:25 +0000 (18:39 -0700)]
Memory-management macros evaluate their arguments multiple times

11 years ago[perl #77496] tied gets scalars and globs confused
Father Chrysostomos [Mon, 25 Oct 2010 01:14:47 +0000 (18:14 -0700)]
[perl #77496] tied gets scalars and globs confused

Make pp_tied use the SvFAKE flag to distinguish between tied scalars
and tied handles. This is now possible as of change 2acc3314.

This fixes the problem of tied($scalar) ignoring the tie if the last
thing returned or assigned happened to be a glob.

11 years agouse compatible types in a conditional expression
Tony Cook [Mon, 25 Oct 2010 00:04:23 +0000 (11:04 +1100)]
use compatible types in a conditional expression

With some compilers NULL is ((void *)0) which isn't type compatible
with the integer returned by hv_exists_ent().

11 years ago[perl #77810] Scalars vs globs
Father Chrysostomos [Sun, 24 Oct 2010 22:50:23 +0000 (15:50 -0700)]
[perl #77810] Scalars vs globs

Stop *{} from returning globs with the SVf_FAKE flag on.

It removes three tests from t/op/gv.t (that I added) that test buggy
edge cases that can no longer occur.

It also modifies tests in t/io/defout.t to keep them passing. I am not
sure that test script serves any purpose any more.

11 years agoUpdate Locale-Maketext to CPAN version 1.17
Todd Rinaldo [Sun, 24 Oct 2010 23:01:31 +0000 (00:01 +0100)]
Update Locale-Maketext to CPAN version 1.17

  [DELTA]

  2010-10-20
    * Release 1.17

    Test release of 1.16_01, versioning even deprecated Guts modules.
    This prevents CPAN upgrade circular heck. Thanks BinGOs

    Tested version fix for CPAN by BinGOs. Bump to stable version and release to public

Signed-off-by: Chris 'BinGOs' Williams <chris@bingosnet.co.uk>
11 years agoUpdate Maintainers.pl with the latest version of Module::CoreList
Chris 'BinGOs' Williams [Sun, 24 Oct 2010 21:02:21 +0000 (22:02 +0100)]
Update Maintainers.pl with the latest version of Module::CoreList

11 years agoIn B.xs, use newSVpvn_flags() instead of sv_newmortal() and sv_setpvn().
Nicholas Clark [Sun, 24 Oct 2010 18:13:10 +0000 (19:13 +0100)]
In B.xs, use newSVpvn_flags() instead of sv_newmortal() and sv_setpvn().

11 years agodon't rely on ghost contexts being unmolested
Zefram [Sun, 24 Oct 2010 13:57:21 +0000 (14:57 +0100)]
don't rely on ghost contexts being unmolested

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.

11 years agoMath::BigRat has blead as its upstream
Florian Ragwitz [Sun, 24 Oct 2010 00:18:25 +0000 (17:18 -0700)]
Math::BigRat has blead as its upstream

11 years agoMake mro_package_moved use a mortal in case mro_isa_changed_in3 croaks
Father Chrysostomos [Sun, 24 Oct 2010 17:42:33 +0000 (10:42 -0700)]
Make mro_package_moved use a mortal in case mro_isa_changed_in3 croaks

11 years agoFix SV leak in Perl_emulate_cop_io(), present since 8b850bd54aa90bd3.
Nicholas Clark [Sun, 24 Oct 2010 16:23:17 +0000 (17:23 +0100)]
Fix SV leak in Perl_emulate_cop_io(), present since 8b850bd54aa90bd3.

11 years agoIn B.xs, tidy up make_*_object().
Nicholas Clark [Sun, 24 Oct 2010 16:02:25 +0000 (17:02 +0100)]
In B.xs, tidy up make_*_object().

All callers to make_temp_object, make_warnings_object and make_cop_io_object
pass in a new mortal, so remove the first argument from all 3 and generate the
mortal within them.

Allow a NULL first argument for make_sv_object - generate a new mortal in this
case.

Ideally we'd remove its first argument too, but currently the output typemap
causes code to be generated that first assigns a new mortal to ST(0), then
passes that to make_sv_object(), and it's not obvious how to trivially fix
that.

11 years agoB::hash() should only work on byte sequences.
Nicholas Clark [Sun, 24 Oct 2010 15:00:23 +0000 (17:00 +0200)]
B::hash() should only work on byte sequences.

11 years agoRefactor B::hash to use Perl_newSVpvf() instead of my_sprintf() and new_SVpvn*.
Nicholas Clark [Sun, 24 Oct 2010 14:23:05 +0000 (16:23 +0200)]
Refactor B::hash to use Perl_newSVpvf() instead of my_sprintf() and new_SVpvn*.

11 years agoMore tests for B::hash().
Nicholas Clark [Sun, 24 Oct 2010 14:05:55 +0000 (16:05 +0200)]
More tests for B::hash().

11 years agoB::cchar can be implemented as an ALIAS of B::cstring.
Nicholas Clark [Sun, 24 Oct 2010 13:51:05 +0000 (15:51 +0200)]
B::cchar can be implemented as an ALIAS of B::cstring.

11 years agoRefactor B::cchar() to create a temporary directly, and use Perl_sv_catpvf().
Nicholas Clark [Sun, 24 Oct 2010 12:23:43 +0000 (14:23 +0200)]
Refactor B::cchar() to create a temporary directly, and use Perl_sv_catpvf().

11 years agoTests for B::cchar().
Nicholas Clark [Sun, 24 Oct 2010 12:07:26 +0000 (14:07 +0200)]
Tests for B::cchar().

11 years agoIn B.xs, refactor cstring() to return a mortal, which simplifies the XS code.
Nicholas Clark [Sun, 24 Oct 2010 11:50:52 +0000 (13:50 +0200)]
In B.xs, refactor cstring() to return a mortal, which simplifies the XS code.

11 years agoTest NUL bytes with B::cstring() and B::perlstring().
Nicholas Clark [Sun, 24 Oct 2010 11:49:12 +0000 (13:49 +0200)]
Test NUL bytes with B::cstring() and B::perlstring().

11 years agoB::perlstring can be implemented as an ALIAS of B::cstring.
Nicholas Clark [Sun, 24 Oct 2010 11:35:06 +0000 (13:35 +0200)]
B::perlstring can be implemented as an ALIAS of B::cstring.

11 years agoIn cstring() in B.xs, use Perl_sv_catpvf(), instead of a temporary buffer.
Nicholas Clark [Sun, 24 Oct 2010 10:26:43 +0000 (12:26 +0200)]
In cstring() in B.xs, use Perl_sv_catpvf(), instead of a temporary buffer.

11 years agoMore comprehensive tests for B::cstring() and B::perlstring().
Nicholas Clark [Sun, 24 Oct 2010 10:13:23 +0000 (12:13 +0200)]
More comprehensive tests for B::cstring() and B::perlstring().

11 years agoIn pp_getlogin, use sv_setpv_mg() to avoid needing an explicit strlen() call.
Nicholas Clark [Sun, 24 Oct 2010 08:59:08 +0000 (09:59 +0100)]
In pp_getlogin, use sv_setpv_mg() to avoid needing an explicit strlen() call.

It was using PUSHp() here, meaning local object code for a call to strlen(),
and for SvSETMAGIC().

11 years agopermit labels to appear before declarations
Zefram [Sat, 16 Oct 2010 21:22:57 +0000 (22:22 +0100)]
permit labels to appear before declarations

Include <label> in productions before <decl> and <package_block>.  This
means that labels can now appear at the beginning of all statement-like
things.  There was no technical reason for the restriction of labels
to substantive statements, and that restriction in any case couldn't be
applied to PLUGSTMT-based plugged-in declarations.

11 years agoPerl_hv_name_add needs to set xhv_name_count in one other place.
Father Chrysostomos [Sun, 24 Oct 2010 01:55:29 +0000 (18:55 -0700)]
Perl_hv_name_add needs to set xhv_name_count in one other place.

Perl_hv_name_add was not setting the name count when upgrading a HEK*
to an array of HEK*s.

11 years ago.gitignore some extra ext/B build products
Tony Cook [Sun, 24 Oct 2010 00:37:25 +0000 (11:37 +1100)]
.gitignore some extra ext/B build products

11 years agoremove ; so the conditional actually controls some code
Tony Cook [Sun, 24 Oct 2010 00:35:02 +0000 (11:35 +1100)]
remove ; so the conditional actually controls some code

11 years agomake the definition of sv_catpv_flags match the declaration
Tony Cook [Sun, 24 Oct 2010 00:26:50 +0000 (11:26 +1100)]
make the definition of sv_catpv_flags match the declaration

This was warning on Win32

11 years agosv_setsv_flags was also using a potentially freed HV
Father Chrysostomos [Sun, 24 Oct 2010 00:04:42 +0000 (17:04 -0700)]
sv_setsv_flags was also using a potentially freed HV

11 years agoTels doesn't want to maintain math stuff anymore
Florian Ragwitz [Sat, 23 Oct 2010 22:50:46 +0000 (15:50 -0700)]
Tels doesn't want to maintain math stuff anymore

11 years agoConvert B to use ExtUtils::Constant::ProxySubs for its constants.
Nicholas Clark [Sat, 23 Oct 2010 18:26:53 +0000 (19:26 +0100)]
Convert B to use ExtUtils::Constant::ProxySubs for its constants.

Previously it was using a Perl script to generate C code with pairs of
 newCONSTSUB(stash,"Foo_BAR",newSViv(Foo_BAR));
 av_push(export_ok,newSVpvn("Foo_BAR",7));
for each constant it exported from C to Perl.

Now it uses a loop to process a table. Object code is size reduced by about 42K
on this system (almost 20%)

11 years agoFor PROXYSUBS, add an option to push all constants' names onto an array.
Nicholas Clark [Sat, 23 Oct 2010 18:00:49 +0000 (19:00 +0100)]
For PROXYSUBS, add an option to push all constants' names onto an array.

Typically this would be 'EXPORT_OK', to add all constants to the @EXPORT_OK
array for this package.

11 years agoFor PROXYSUBS, allow an explicit XS_SUBNAME of undef to suppress its generation.
Nicholas Clark [Sat, 23 Oct 2010 17:41:04 +0000 (18:41 +0100)]
For PROXYSUBS, allow an explicit XS_SUBNAME of undef to suppress its generation.

11 years agoChange concise.t to use Storable to test stash/src rendering of constant subs.
Nicholas Clark [Sat, 23 Oct 2010 17:06:42 +0000 (18:06 +0100)]
Change concise.t to use Storable to test stash/src rendering of constant subs.

This allows B to be refactored to change the implementation of its constants.
Storable strives to main compatibility with 5.004 (or so), so unlikely to be
refactored soon.

11 years agoS_glob_assign_glob was using a potentially freed HV
Father Chrysostomos [Sat, 23 Oct 2010 16:48:31 +0000 (09:48 -0700)]
S_glob_assign_glob was using a potentially freed HV

11 years agoperlipc: Convert tabs to spaces.
Shlomi Fish [Wed, 21 Jul 2010 18:31:50 +0000 (21:31 +0300)]
perlipc: Convert tabs to spaces.

Signed-off-by: David Golden <dagolden@cpan.org>
11 years agoperlipc: fix bug in example
David Golden [Sat, 23 Oct 2010 11:13:49 +0000 (07:13 -0400)]
perlipc: fix bug in example

11 years agoAdd functions for adding and deleting stash names
Father Chrysostomos [Sat, 23 Oct 2010 06:56:29 +0000 (23:56 -0700)]
Add functions for adding and deleting stash names

11 years agoFor PROXYSUBS, also avoid calling get_missing_hash() in the XS constant code.
Nicholas Clark [Sat, 23 Oct 2010 06:36:53 +0000 (07:36 +0100)]
For PROXYSUBS, also avoid calling get_missing_hash() in the XS constant code.

(Follow on from 4639bd9c727433ad)
For modules such as File::Glob and I18N::Langinfo, which always define all
constants on all platforms, this saves about 700 bytes of object code.

11 years agoClarify and correct description of comma operator in scalar context
Ian Goodacre [Sat, 23 Oct 2010 04:42:55 +0000 (17:42 +1300)]
Clarify and correct description of comma operator in scalar context

The guarantee that in scalar context the comma operator evaluates its
arguments in scalar context is overstated.

In perl 5.10.0

print "Scalar assignment:\n";
$x = ( context(1), context(2), context(3) );

print "Scalar assignment in sub:\n";
sub list { ( context(1), context(2), context(3) ) }
$x = list();

sub context {

    if(wantarray) {
        print "list context\n";
    } elsif(defined(wantarray)) {
        print "scalar context\n";
    } else {
        print "void context\n";
    }
}

prints:

scalar assignment:
void context
void context
scalar context
Scalar assignment in sub:
scalar context
scalar context
scalar context

This leaves only the right argument of the last comma operator in a list as
the only one that might always be evaluated in scalar context.

The comments on the sample outputs were at best ambiguous if not misleading
or false, and also unnecessarily pejorative of perl4. The revised comments
less ambiguously refer to the last expression in the list (@y in the example)
rather than to the literal list that is the argument of the assignment
operator.

11 years agoCorrected Perl5 example of Formatted output and significant digits
Ian Goodacre [Sat, 23 Oct 2010 03:46:02 +0000 (16:46 +1300)]
Corrected Perl5 example of Formatted output and significant digits

The following note says "Your results may vary...", so maybe some perls
did give the value in the example as it was, but perl v5.10.0 on Intel
Pentium based system gives the more accurate revised output.

I am also dubious about the values given for perl4 as the error is quite
large, but I don't have perl4 easily available to test.

11 years agoAdded missing variable name in example code
Ian Goodacre [Sat, 23 Oct 2010 00:38:52 +0000 (13:38 +1300)]
Added missing variable name in example code

11 years agoRemove cruft from B's Makefile.PL
Nicholas Clark [Fri, 22 Oct 2010 20:10:15 +0000 (21:10 +0100)]
Remove cruft from B's Makefile.PL

$noecho is unused since 70c68212ed57a5a0 (in 2005)
$e and $exeout_flag are unused since e8edd1e67bd80dbb (in 1999)
All the clean FILES apart from 'defsubs.h' are vestigial or duplicate defaults.
The rule for LIBS created by post_constants() seems superfluous.

11 years agoFor PROXYSUBS, get the hash from the fetch to pass to the subsequent store.
Nicholas Clark [Fri, 22 Oct 2010 17:59:36 +0000 (19:59 +0200)]
For PROXYSUBS, get the hash from the fetch to pass to the subsequent store.

This saves the store having to (re)compute the hash.

11 years agoFor PROXYSUBS, loop with do {} not while, as there is always at least 1 entry.
Nicholas Clark [Fri, 22 Oct 2010 14:58:36 +0000 (15:58 +0100)]
For PROXYSUBS, loop with do {} not while, as there is always at least 1 entry.

11 years agoFor PROXYSUBS, avoid calling get_missing_hash() if all symbols are present.
Nicholas Clark [Fri, 22 Oct 2010 12:03:09 +0000 (14:03 +0200)]
For PROXYSUBS, avoid calling get_missing_hash() if all symbols are present.

For modules such as File::Glob and I18N::Langinfo, which always define all
constants on all platforms, this saves about 1K of object code, plus 1 (empty)
anonymous hash (per process or thread) in
%ExtUtils::Constant::ProxySubs::Missing.

11 years agoFor PROXYSUBS, give the notfound struct file scope.
Nicholas Clark [Fri, 22 Oct 2010 10:15:17 +0000 (12:15 +0200)]
For PROXYSUBS, give the notfound struct file scope.

All others remain within the scope of the BOOT block.

11 years agoMinor refactor of ExtUtils::Constant::ProxySubs.
Nicholas Clark [Fri, 22 Oct 2010 08:36:44 +0000 (10:36 +0200)]
Minor refactor of ExtUtils::Constant::ProxySubs.

Use a lexical in place of repeated lc()'s in WriteConstants.
Require ExtUtils::Constant::ProxySubs early, to avoid leaving empty files if it
fails for any reason.

11 years agoFix on doc patch from Moritz Lenz (from IRC)
H.Merijn Brand [Fri, 22 Oct 2010 10:40:26 +0000 (12:40 +0200)]
Fix on doc patch from Moritz Lenz (from IRC)

11 years agoMerge the implementation of B::CV::{ROOT,START}.
Nicholas Clark [Fri, 22 Oct 2010 07:47:09 +0000 (08:47 +0100)]
Merge the implementation of B::CV::{ROOT,START}.

11 years agoNote the U8-sized space in intrpvar.h created by 737c24fc85ea612b.
Nicholas Clark [Fri, 22 Oct 2010 07:37:29 +0000 (08:37 +0100)]
Note the U8-sized space in intrpvar.h created by 737c24fc85ea612b.

11 years agoBetter diagnostics for XS::APItest::test_cv_getset_call_checker().
Nicholas Clark [Fri, 22 Oct 2010 07:24:18 +0000 (08:24 +0100)]
Better diagnostics for XS::APItest::test_cv_getset_call_checker().

11 years agoS_tokeq()'s fast scan loop should terminate on \\ not \
Nicholas Clark [Fri, 22 Oct 2010 06:41:39 +0000 (07:41 +0100)]
S_tokeq()'s fast scan loop should terminate on \\ not \

As-was, it would drop out of the scanner into the backslashed-backslash
processing loop earlier than need be, and hence would be copying the octets
of strings (in place) as soon as any backslash had been seen. Now it defers
copying until copying is actually unavoidable.

11 years agoperldelta: fix typo and POD syntax
Father Chrysostomos [Fri, 22 Oct 2010 04:49:33 +0000 (21:49 -0700)]
perldelta: fix typo and POD syntax

11 years agoperldelta for things I committed recently
Father Chrysostomos [Fri, 22 Oct 2010 04:42:35 +0000 (21:42 -0700)]
perldelta for things I committed recently

11 years agoBump re’s version
Father Chrysostomos [Fri, 22 Oct 2010 04:41:47 +0000 (21:41 -0700)]
Bump re’s version

11 years agoBump charnames’ version
Father Chrysostomos [Fri, 22 Oct 2010 03:42:47 +0000 (20:42 -0700)]
Bump charnames’ version

11 years ago[perl #78488] Bleadperl 304474c3 breaks GFUJI/Test-LeakTrace-0.13.tar.gz
Father Chrysostomos [Fri, 22 Oct 2010 01:18:42 +0000 (18:18 -0700)]
[perl #78488] Bleadperl 304474c3 breaks GFUJI/Test-LeakTrace-0.13.tar.gz

This commit restores an SvREFCNT_dec that was inadvertently removed.

c8bbf67 removed the SvREFCNT_dec(HeVAL(entry)), adding
SvREFCNT_dec(old_val) instead.

304474c3 reverted that block, but failed to restore the
SvREFCNT_dec(HeVAL(entry)).

The result was that the %INC entry created by do "file" was leaking.

(Y’know, giving out commit bits to just *anyone* who comes along
with a few patches seems dangerous. Has anybody been looking at what
I’m doing?)

11 years agoFix calling conventions in malloc_ctl.h
Jan Dubois [Fri, 22 Oct 2010 01:00:14 +0000 (18:00 -0700)]
Fix calling conventions in malloc_ctl.h

The Perl_malloc() etc. functions are *also* declared in
proto.h, so the declarations need to match.  The inclusion
of malloc_ctl.h into perl.h had to be moved down until
after the point where PERL_CALLCONV was completely defined.

11 years agoregcomp.c: /d not overriding /u
Karl Williamson [Wed, 20 Oct 2010 19:21:04 +0000 (13:21 -0600)]
regcomp.c: /d not overriding /u

The setting of the charset regex modifiers was wrong.  /d didn't
override /u nor /l, and similarly /u and /l didn't properly override
each other.

11 years agoregcomp.c: Fix typo in comment
Karl Williamson [Sun, 17 Oct 2010 16:38:37 +0000 (10:38 -0600)]
regcomp.c: Fix typo in comment

11 years agoFix ext/XS-APItest/t/multicall.t warning
Jerry D. Hedden [Thu, 21 Oct 2010 21:47:50 +0000 (14:47 -0700)]
Fix ext/XS-APItest/t/multicall.t warning

'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.