fork crashes on perl 5.10.0/Win32 (when using Variable::Magic) #9757
Comments
From @kmxHi, the following script: *** crashes during "fork" call on win32/strawberry perl 5.10.0.4 & 5.10.0.5 On Win32 Active state perl 5.10 the program (fork call) does not work The same script works fine on win32/strawberry perl 5.8.9.0 & 5.8.9.1 - To be honest I am not sure if this is a bug in Variable::Magic or perl Here is some debug info about the crash point: (gdb) run test.pl Perl Info |
From bitcard@profvince.com
For kmx: cc'ing the progress to the p5p list. For p5p: This is a cross report of I think this can be fixed by making PerlProcFork() call Does someone know why perl_clone_using() is called here without this Vincent. |
|
The RT System itself - Status changed from 'new' to 'open' |
From @kmxHi, could you please confirm or reject this issue as a bug in fork I would appreciate if somebody could test whether the same behaviour This simple script kills Win32 perl 5.10.0 use Variable::Magic 'wizard'; -- |
From @kmxHi, I have investigated this issue further and found out that it is present Based on Vicent's post above I have tried this patch to win32/perlhost.h - PerlInterpreter *new_perl = perl_clone_using((PerlInterpreter*)aTHX, 1, and IT SOLVES the described fork issue but I am not able to say that I would appreciate any feedback. -- |
From @csjewellCan I request that this fix (or a better-implemented fix for this bug) I'm given to understand (talk to kmx about that to be sure) that this --Curtis Jewell Mon, 03 Aug 2009 07:03 -0700, "kmx via RT" <perlbug-followup@perl.org>
%DCL-E-MEM-BAD, bad memory [I use PC-Alpine, which deliberately does not display colors and pictures in HTML mail] |
From @iabynOn Tue, Aug 04, 2009 at 03:49:29PM -0600, Curtis Jewell wrote:
I would be extremely reluctant to put this in RC1, and if it's not in RC1,
So if I understand it correctly, Variable::Magic has been made to work The thing that puzzles me is that is doesn't explode on 5.8.x. -- |
From @kmx
-- |
From @csjewellOn Tue, 04 Aug 2009 23:46 +0100, "Dave Mitchell" <davem@iabyn.com>
I can understand that. I wouldn't want to take a patch that low-level in And I have to admit that I would prefer that Catalyst not have the Would it help if I ran a smoker over perl with that patch, and if so, --Curtis %DCL-E-MEM-BAD, bad memory [I use PC-Alpine, which deliberately does not display colors and pictures in HTML mail] |
From @csjewell(forgot to use my outbound to-list address when sending it out ----- Original message ----- On Tue, 04 Aug 2009 23:46 +0100, "Dave Mitchell" <davem@iabyn.com>
I can understand that. I wouldn't want to take a patch that low-level in And I have to admit that I would prefer that Catalyst not have the Would it help if I ran a smoker over perl with that patch, and if so, --Curtis %DCL-E-MEM-BAD, bad memory [I use PC-Alpine, which deliberately does not display colors and -- %DCL-E-MEM-BAD, bad memory [I use PC-Alpine, which deliberately does not display colors and pictures in HTML mail] |
From perl@profvince.com
Yes, exactly. Variable::Magic stores coderefs in a thread local context.
I've no idea about why it works with 5.8, nor why the pointer table is Vincent. |
From perl@profvince.comCC-ing the list.
Yes, exactly. Variable::Magic stores coderefs in a thread local context.
I've no idea about why it works with 5.8, nor why the pointer table is Vincent. |
From @iabynOn Wed, Aug 05, 2009 at 03:16:05PM +0200, Vincent Pit wrote:
That was a change to threads.xs a while ago. Originally neither kept the Anyway, I've caved in and applied the following to commits (the first is commit 46a76da win32/perlhost.h: use symbolic constant Affected files ... Differences ... Inline Patchdiff --git a/win32/perlhost.h b/win32/perlhost.h
index 6e3fcd2..c2473c9 100644
--- a/win32/perlhost.h
+++ b/win32/perlhost.h
@@ -1835,7 +1835,8 @@ PerlProcFork(struct IPerlProc* piPerl)
return -1;
}
h = new CPerlHost(*(CPerlHost*)w32_internal_host);
- PerlInterpreter *new_perl = perl_clone_using((PerlInterpreter*)aTHX, 1,
+ PerlInterpreter *new_perl = perl_clone_using((PerlInterpreter*)aTHX,
+ CLONEf_COPY_STACKS,
h->m_pHostperlMem,
h->m_pHostperlMemShared,
h->m_pHostperlMemParse,
add CLONEf_KEEP_PTR_TABLE to win fork emulation. Affected files ... Differences ... Inline Patchdiff --git a/win32/perlhost.h b/win32/perlhost.h
index c2473c9..c02d409 100644
--- a/win32/perlhost.h
+++ b/win32/perlhost.h
@@ -1836,7 +1836,8 @@ PerlProcFork(struct IPerlProc* piPerl)
}
h = new CPerlHost(*(CPerlHost*)w32_internal_host);
PerlInterpreter *new_perl = perl_clone_using((PerlInterpreter*)aTHX,
- CLONEf_COPY_STACKS,
+ CLONEf_COPY_STACKS
+ | CLONEf_KEEP_PTR_TABLE,
h->m_pHostperlMem,
h->m_pHostperlMemShared,
h->m_pHostperlMemParse,
-- Now is the discount of our winter tent |
From perl@profvince.comThanks for this.
Now I know why it doesn't crash on 5.8 : I explicitely disabled
threads.pm manually destroys the pointer table after the new interpreter Vincent. |
From @iabynOn Wed, Aug 05, 2009 at 04:02:54PM +0200, Vincent Pit wrote:
Ah, the penny drops! The real issue is that in perl_clone_using(), it frees the ptr_table With b0b93b3, I've fixed I'll be pushing this into maint in a moment. I wonder: for existing 5.10.0 users, whether Variable::Magic could -- |
From @kmxHi,
I have build perl/Win32 (using mingw) from blead with the Many thanks to all of you involved in solving this issues.
-- |
From @iabynOn Wed, Aug 05, 2009 at 03:33:34PM +0100, Dave Mitchell wrote:
PS can someone confirm ASAP that GitLive-maint-5.10-1727-g02784d7 or later Thanks. -- |
From @csjewellOn Wed Aug 05 14:01:26 2009, davem wrote:
Can you live with ASAP == between 12 and 24 hours? I'm not at the (speaking of which, I ran a 4-build gcc smoker (-DDEBUGGING and --Curtis |
From @iabynOn Wed, Aug 05, 2009 at 12:42:12PM -0700, kmx via RT wrote:
Yes, it'll be in 5.10.1 -- |
From @csjewellOn Thu, 06 Aug 2009 16:48 +0100, "Dave Mitchell" <davem@iabyn.com>
and it works in the Perl that was tagged 5.10.1 RC1: C:\NewFolder>\perl\bin\perl.exe -e "use namespace::clean; fork; print (test.pl being the script used in perl rt# 66158) C:\NewFolder>\perl\bin\perl test.pl The Perl in question? (I just compiled this from Summary of my perl5 (revision 5 version 10 subversion 1 patch maint-5.10 Characteristics of this binary (from libperl): Automated smoke report for 5.10.1 patch Summary: PASS O = OK F = Failure(s), extended report at the bottom blead-1787-ga0db33f Configuration (common) none Locally applied patches: Compiler messages(gcc): -- %DCL-E-MEM-BAD, bad memory [I use PC-Alpine, which deliberately does not display colors and pictures in HTML mail] |
From @csjewellIf I didn't say so earlier, thanks for getting this fix into 5.10.1 to %DCL-E-MEM-BAD, bad memory [I use PC-Alpine, which deliberately does not display colors and pictures in HTML mail] |
From @demerphq2009/8/6 Curtis Jewell <lists.perl.perl5-porters@csjewell.fastmail.us>:
Hrm, this is a bug by me. I have to fix that. (Note the tag doesnt Yves -- |
From @demerphq2009/8/6 demerphq <demerphq@gmail.com>:
Fixed. It should have said: maint-5.10 2009-08-06.00:19:12 Yves -- |
From @csjewellOn Thu, 06 Aug 2009 19:04 +0200, "demerphq" <demerphq@gmail.com> wrote:
I thought the tag at the end looked funny. (I did download it BEFORE it --Curtis %DCL-E-MEM-BAD, bad memory [I use PC-Alpine, which deliberately does not display colors and pictures in HTML mail] |
From @kmx
Once again many thanks for this "last-minute-before-5.10.1-patch" to all -- |
From @demerphq2009/8/6 Curtis Jewell <lists.perl.perl5-porters@csjewell.fastmail.us>:
Yes right, that was the bug. Yves -- |
From @iabynNow fixed in 5.10.1 |
|
@iabyn - Status changed from 'open' to 'resolved' |
Migrated from rt.perl.org#66158 (status was 'resolved')
Searchable as RT66158$
The text was updated successfully, but these errors were encountered: