fork.t fails on Win32 since v5.15.4-465-g676a678 #11927
Comments
From @tonycozCreated by @tonycozx64 MSVC build, but smokes have been failing too: C:\Users\tony\dev\perl\git\perl\t>.\perl harness -v op/fork.t # Failed at ./test.pl line 1076 Test Summary Report Tested to be broken by 676a678: C:\Users\tony\dev\perl\git\perl\t>git log -n1 narrower localisation of PL_compcv around eval PL_compcv used to be localised around the entire string eval process, Similar to the fix there, this change moves the localisation of PL_compcv C:\Users\tony\dev\perl\git\perl\t>.\perl harness -v op/fork.t # Failed at ./test.pl line 1033 Test Summary Report C:\Users\tony\dev\perl\git\perl\t>cd .. C:\Users\tony\dev\perl\git\perl>git clean -dxf C:\Users\tony\dev\perl\git\perl>git checkout "676a678ac0683a727a07d56ed1a6e1fb59 C:\Users\tony\dev\perl\git\perl>cd win32 C:\Users\tony\dev\perl\git\perl\win32>dmake test-prep C:\Users\tony\dev\perl\git\perl\t>.\perl harness -v op/fork.t Perl Info |
From @tonycozA little more information: C:\Users\tony\dev\perl\git\perl>.\perl -Ilib ..\forktest.pl C:\Users\tony\dev\perl\git\perl>type ..\forktest.pl |
|
@tonycoz - Status changed from 'new' to 'open' |
From @cpansproutFixed in commit 44c19e2. -- Father Chrysostomos |
From [Unknown Contact. See original ticket]Fixed in commit 44c19e2. -- Father Chrysostomos |
|
@cpansprout - Status changed from 'open' to 'resolved' |
From [email protected]Father Chrysostomos via RT wrote:
Is it really? I know I fixed *a* bug there, and I found it because I -zefram |
From @cpansproutOn Sat Feb 18 14:52:55 2012, zefram@fysh.org wrote:
I that case I misunderstood your commit message. I’ll re-open it. -- Father Chrysostomos |
|
@cpansprout - Status changed from 'resolved' to 'open' |
From @tonycozOn Fri Feb 03 00:49:17 2012, tonyc wrote:
I should have tried this before: C:\Users\tony\dev\perl\git\perl>.\perl -Dl -Ilib ..\forktest.pl EXECUTING... (..\forktest.pl:0) Entering new RUNOPS level |
From @janduboisSorry, I didn't get around to looking at it yet, but here are The crash happens in win32/perlhost.c on line 1727: 1722: EAX = 00000002 EBX = 0014014C ECX = 00000000 00000008 = ???????? I have no idea how this might relate to the change triggering the Cheers, |
From [email protected]I have distilled this down to: .\perl -Dl -e "eval q{fork}" EXECUTING... (-e:0) Entering new RUNOPS level |
From @icerider70On Sat, Mar 31, 2012 at 11:27 PM, Alex Gough <alex@earth.li> wrote:
<output showing panic snipped> For comparison, here's the output of the same command when run against a EXECUTING... (-e:0) Entering new RUNOPS level |
From @khwilliamsonAttached is a diff file between the before/after with xterm color |
From @khwilliamsonwin32.diff EXECUTING...
(-e:0) Entering new RUNOPS level
4,9c4,9
< (-e:0) ENTER "block" scope 2 (savestack=0) at ..\pp_ctl.c:�[30;46m2116�[0m
< (-e:0) CX 0 PUSH BLOCK (scope 2,2) at ..\pp_ctl.c:�[30;46m2119�[0m
< (-e:1) ENTER "eval" scope 3 (savestack=3) at ..\pp_ctl.c:�[30;46m4208�[0m
< (-e:1) CX 1 PUSH EVAL (scope 3,3) at ..\pp_ctl.c:�[30;46m4239�[0m
< (�[30;46m(eval 1)�[0m:1) �[30;46m �[0mENTER "evalcomp" scope 4 (savestack=�[30;46m48�[0m) at ..\pp_ctl.c:�[30;46m3528�[0m
< ((eval 1):1) ENTER scope 5 (savestack=53) at ..\perly.c:�[30;46m362�[0m
---
> (-e:0) ENTER "block" scope 2 (savestack=0) at ..\pp_ctl.c:�[30;43m2135�[0m
> (-e:0) CX 0 PUSH BLOCK (scope 2,2) at ..\pp_ctl.c:�[30;43m2138�[0m
> (-e:1) ENTER "eval" scope 3 (savestack=3) at ..\pp_ctl.c:�[30;43m4288�[0m
> (-e:1) CX 1 PUSH EVAL (scope 3,3) at ..\pp_ctl.c:�[30;43m4319�[0m
> (�[30;43m-e�[0m:1) ENTER "evalcomp" scope 4 (savestack=�[30;43m45�[0m) at ..\pp_ctl.c:�[30;43m3594�[0m
> ((eval 1):1) ENTER scope 5 (savestack=53) at ..\perly.c:�[30;43m358�[0m
((eval 1):2) savestack: releasing items 89 -> 73
11,17c11,19
< ((eval 1):2) ENTER scope 6 (savestack=�[30;46m73�[0m) at op.c:�[30;46m9869�[0m
< ((eval 1):2) LEAVE scope �[30;46m6�[0m (savestack=�[30;46m78�[0m) at op.c:�[30;46m10284�[0m
< ((eval 1):2) savestack: releasing items �[30;46m78�[0m -> �[30;46m73�[0m
< ((eval 1):2) ENTER scope �[30;46m6�[0m (savestack=�[30;46m73�[0m) at op.c:�[30;46m1462�[0m
< ((eval 1):2) LEAVE scope �[30;46m6�[0m (savestack=�[30;46m76�[0m) at op.c:�[30;46m1467�[0m
< ((eval 1):2) savestack: releasing items �[30;46m76�[0m -> �[30;46m73�[0m
< ((eval 1):2) LEAVE scope 5 (savestack=73) at ..\perly.c:�[30;46m723�[0m
---
> ((eval 1):2) ENTER scope 6 (savestack=�[30;43m75) at op.c:2797�[0m
> �[30;43m((eval 1):2) ENTER scope 7 (savestack=75�[0m) at op.c:�[30;43m9969�[0m
> ((eval 1):2) LEAVE scope �[30;43m7�[0m (savestack=�[30;43m80�[0m) at op.c:�[30;43m10420�[0m
> ((eval 1):2) savestack: releasing items �[30;43m80�[0m -> �[30;43m75�[0m
> ((eval 1):2) ENTER scope �[30;43m7�[0m (savestack=�[30;43m75�[0m) at op.c:�[30;43m1501�[0m
> ((eval 1):2) LEAVE scope �[30;43m7�[0m (savestack=�[30;43m78�[0m) at op.c:�[30;43m1506�[0m
> ((eval 1):2) savestack: releasing items �[30;43m78�[0m -> �[30;43m75�[0m
> �[30;43m((eval 1):2) LEAVE scope 6 (savestack=75) at op.c:2800�[0m
> ((eval 1):2) LEAVE scope 5 (savestack=73) at ..\perly.c:�[30;43m719�[0m
((eval 1):2) savestack: releasing items 73 -> 53
19,35c21,27
< ((eval 1):2) LEAVE "evalcomp" scope 4 (savestack=53) at ..\pp_ctl.c:�[30;46m3577�[0m
< ((eval 1):2) savestack: releasing items 53 -> �[30;46m48�[0m
< ((eval 1):2) CX 1 POP EVAL (scope 3,3) at ..\pp_ctl.c:�[30;46m4297�[0m
< (-e:1) LEAVE "eval" scope 3 (savestack=�[30;46m50�[0m) at ..\pp_ctl.c:�[30;46m4325�[0m
< (-e:1) savestack: releasing items �[30;46m50�[0m -> 3
< (�[30;46m(eval 1):2) JUMPENV_PUSH level=1 at perlhost.h:1739�[0m
< �[30;46m((eval 1):2) Entering new RUNOPS level�[0m
< �[30;46m((eval 1):2) CX 1 POP EVAL (scope 3,3) at ..\pp_ctl.c:4297�[0m
< �[30;46m(-e:1) LEAVE "eval" scope 3 (savestack=50) at ..\pp_ctl.c:4325�[0m
< �[30;46m(-e:1) savestack: releasing items 50 -> 3�[0m
< �[30;46m(-e:1) CX 0 POP BLOCK (scope 2,2) at ..\pp_ctl.c:2137�[0m
< �[30;46m((eval 1):0) LEAVE "block" scope 2 (savestack=3) at ..\pp_ctl.c:2145�[0m
< �[30;46m((eval 1):0) savestack: releasing items 3 -> 0�[0m
< �[30;46m((eval 1):0) leaving RUNOPS level�[0m
< �[30;46m((eval 1):0) JUMPENV_POP level=1 at perlhost.h:1780�[0m
< �[30;46m(�[0m-e:1) CX 0 POP BLOCK (scope 2,2) at ..\pp_ctl.c:�[30;46m2137�[0m
< (-e:0) LEAVE "block" scope 2 (savestack=3) at ..\pp_ctl.c:�[30;46m2145�[0m
---
> ((eval 1):2) LEAVE "evalcomp" scope 4 (savestack=53) at ..\pp_ctl.c:�[30;43m3722�[0m
> ((eval 1):2) savestack: releasing items 53 -> �[30;43m45�[0m
> ((eval 1):2) CX 1 POP EVAL (scope 3,3) at ..\pp_ctl.c:�[30;43m4378�[0m
> (-e:1) LEAVE "eval" scope 3 (savestack=�[30;43m47�[0m) at ..\pp_ctl.c:�[30;43m4407�[0m
> (-e:1) savestack: releasing items �[30;43m47�[0m -> 3
> (-e:1) CX 0 POP BLOCK (scope 2,2) at ..\pp_ctl.c:�[30;43m2156�[0m
> (-e:0) LEAVE "block" scope 2 (savestack=3) at ..\pp_ctl.c:�[30;43m2164�[0m
(-e:0) savestack: releasing items 3 -> 0
(-e:0) leaving RUNOPS level
38,40c30,37
< (-e:0) JUMPENV_JUMP(2) level=1 at perl.c:�[30;46m4920�[0m
< (-e:0) JUMPENV_POP level=1 at perl.c:�[30;46m2330�[0m
< (-e:�[30;46m0�[0m) LEAVE scope �[30;46m1�[0m (savestack=�[30;46m0�[0m) at �[30;46mperl�[0m.c:�[30;46m570�[0m
---
> (-e:0) JUMPENV_JUMP(2) level=1 at perl.c:�[30;43m4951�[0m
> (-e:0) JUMPENV_POP level=1 at perl.c:�[30;43m2347�[0m
> �[30;43m((eval 1):2) (dopoptoeval(): found eval at cx=1)�[0m
> �[30;43m((eval 1):2) CX 1 POP EVAL (scope 3,3) at ..\pp_ctl.c:1747�[0m
> (-e:�[30;43m1�[0m) LEAVE scope �[30;43m3�[0m (savestack=�[30;43m47�[0m) at �[30;43m..\pp_ctl.c:1765�[0m
> �[30;43m(-e:1) savestack: releasing items 47 -> 3�[0m
> �[30;43m(-e:1) JUMPENV_JUMP(3) level=0 at ..\pp_ctl�[0m.c:�[30;43m1795�[0m
> �[30;43mpanic: top_env, v=3�[0m
|
From @bulk88I did some thinking on this bug. I made a macro expanded Perl_die_unwind callstack of **CHILD** OS thread, the die_unwind is my version, so line
______________________________________________________ It seems to me the bug is that this line (look in the die_unwind.c Debugger shows this. - Itop_env 0x00908044 {je_prev=0x00000000 {je_prev=??? je_buf=0x00000004 I'll guess the longjmp info is never collected. Platform: Characteristics of this binary (from libperl): |
From @bulk88void
Perl_die_unwind (PerlInterpreter * my_perl, SV * msv)
{
extern int Perl___notused (void);
SV *exceptsv = Perl_sv_mortalcopy (my_perl, msv);
U8 in_eval = (my_perl->Iin_eval);
(void) ((msv) || (_assert ("msv", "..\\pp_ctl.c", 1676), 0));
if (in_eval)
{
I32 cxix;
I32 gimme;
/*
* Historically, perl used to set ERRSV ($@) early in the die
* process and rely on it not getting clobbered during unwinding.
* That sucked, because it was liable to get clobbered, so the
* setting of ERRSV used to emit the exception from eval{} has
* been moved to much later, after unwinding (see just before
* JMPENV_JUMP below). However, some modules were relying on the
* early setting, by examining $@ during unwinding to use it as
* a flag indicating whether the current unwinding was caused by
* an exception. It was never a reliable flag for that purpose,
* being totally open to false positives even without actual
* clobberage, but was useful enough for production code to
* semantically rely on it.
*
* We'd like to have a proper introspective interface that
* explicitly describes the reason for whatever unwinding
* operations are currently in progress, so that those modules
* work reliably and $@ isn't further overloaded. But we don't
* have one yet. In its absence, as a stopgap measure, ERRSV is
* now *additionally* set here, before unwinding, to serve as the
* (unreliable) flag that it used to.
*
* This behaviour is temporary, and should be removed when a
* proper way to detect exceptional unwinding has been developed.
* As of 2010-12, the authors of modules relying on the hack
* are aware of the issue, because the modules failed on
* perls 5.13.{1..7} which had late setting of $@ without this
* early-setting hack.
*/
if (!(in_eval & 4))
{
((exceptsv)->sv_flags &= ~0x00080000);
Perl_sv_setsv_flags (my_perl,
(*
((0 +
((my_perl->Ierrgv))->sv_u.svu_gp)->gp_sv ? &((0 +
((my_perl->Ierrgv))->
sv_u.svu_gp)->
gp_sv) : &((0 +
(Perl_gv_add_by_type
(my_perl,
((my_perl->Ierrgv)), SVt_NULL))->sv_u.svu_gp)->gp_sv))), exceptsv, 2 | 512 | 1024);
}
while ((cxix = S_dopoptoeval (my_perl, ((my_perl->Icurstackinfo)->si_cxix))) < 0
&& (my_perl->Icurstackinfo)->si_prev)
{
Perl_dounwind (my_perl, -1);
do
{
SV **sp = (my_perl->Istack_sp);
PERL_SI *const prev = (my_perl->Icurstackinfo)->si_prev;
if (((my_perl->Idebug) & 0x00000004))
{
int i = -1;
PERL_SI *p = (my_perl->Icurstackinfo);
while (p)
{
i++;
p = p->si_prev;
}
Perl_deb (my_perl, "pop STACKINFO %d at %s:%d\n", i, "..\\pp_ctl.c", 1720);
}
if (!prev)
{
PerlIO_printf (((my_perl->Istderrgv)
&& (((svtype) (((my_perl->Istderrgv))->sv_flags & 0xff)) ==
SVt_PVGV)
&& ((0 + ((my_perl->Istderrgv))->sv_u.svu_gp)->gp_io)
&&
((XPVIO
*) (((0 +
((my_perl->Istderrgv))->sv_u.svu_gp)->gp_io))->sv_any)->
xio_ofp
? ((XPVIO
*) (((0 +
((my_perl->Istderrgv))->sv_u.svu_gp)->gp_io))->
sv_any)->xio_ofp : Perl_PerlIO_stderr (my_perl)),
"panic: POPSTACK\n");
Perl_my_exit (my_perl, 1);
}
do
{
((XPVAV *) ((my_perl->Icurstack))->sv_any)->xav_fill =
sp - (my_perl->Istack_base);
(my_perl->Istack_base) = ((prev->si_stack)->sv_u.svu_array);
(my_perl->Istack_max) =
(my_perl->Istack_base) + ((XPVAV *) (prev->si_stack)->sv_any)->xav_max;
sp = (my_perl->Istack_sp) =
(my_perl->Istack_base) + ((XPVAV *) (prev->si_stack)->sv_any)->xav_fill;
(my_perl->Icurstack) = prev->si_stack;
}
while (0);
(my_perl->Icurstackinfo) = prev;
}
while (0);
}
if (cxix >= 0)
{
I32 optype;
SV *namesv;
PERL_CONTEXT *cx;
SV **newsp;
COP *oldcop;
JMPENV *restartjmpenv;
OP *restartop;
if (cxix < ((my_perl->Icurstackinfo)->si_cxix))
Perl_dounwind (my_perl, cxix);
if (((my_perl->Idebug) & 0x00000004))
Perl_deb (my_perl, "CX %ld %s %s (scope %ld,%ld) at %s:%d\n",
(long) ((my_perl->Icurstackinfo)->si_cxix), "POP",
PL_block_type[((&((my_perl->Icurstackinfo)->si_cxstack)
[((my_perl->Icurstackinfo)->si_cxix)])->cx_u.cx_subst.
sbu_type & 0xf)], (long) (my_perl->Iscopestack_ix),
(long) (((my_perl->Icurstackinfo)->
si_cxstack)[((my_perl->Icurstackinfo)->si_cxix)].cx_u.cx_blk.
blku_oldscopesp), "..\\pp_ctl.c", 1735);;
cx = &((my_perl->Icurstackinfo)->si_cxstack)[((my_perl->Icurstackinfo)->si_cxix)--], newsp =
(my_perl->Istack_base) + cx->cx_u.cx_blk.blku_oldsp, (my_perl->Icurcop) =
cx->cx_u.cx_blk.blku_oldcop, (my_perl->Imarkstack_ptr) =
(my_perl->Imarkstack) + cx->cx_u.cx_blk.blku_oldmarksp, (my_perl->Iscopestack_ix) =
cx->cx_u.cx_blk.blku_oldscopesp, (my_perl->Icurpm) = cx->cx_u.cx_blk.blku_oldpm, gimme =
cx->cx_u.cx_blk.blku_gimme;;
if (((cx)->cx_u.cx_subst.sbu_type & 0xf) != 10)
{
STRLEN msglen;
const char *message =
(((my_perl->ISv) =
(exceptsv)),
((((my_perl->ISv))->sv_flags & (0x00000400)) ==
0x00000400 ? ((msglen = ((XPV *) ((my_perl->ISv))->sv_any)->xpv_cur),
((const char *) (0 +
((my_perl->ISv))->sv_u.
svu_pv))) : (const char *)
Perl_sv_2pv_flags (my_perl, (my_perl->ISv), &msglen, 2 | 32)));
Perl_PerlIO_write (my_perl,
((my_perl->Istderrgv)
&& (((svtype) (((my_perl->Istderrgv))->sv_flags & 0xff)) ==
SVt_PVGV)
&& ((0 + ((my_perl->Istderrgv))->sv_u.svu_gp)->gp_io)
&&
((XPVIO *) (((0 + ((my_perl->Istderrgv))->sv_u.svu_gp)->gp_io))->
sv_any)->
xio_ofp
? ((XPVIO
*) (((0 +
((my_perl->Istderrgv))->sv_u.svu_gp)->gp_io))->sv_any)->
xio_ofp : Perl_PerlIO_stderr (my_perl)),
(const char *) "panic: die ", 11);
Perl_PerlIO_write (my_perl,
((my_perl->Istderrgv)
&& (((svtype) (((my_perl->Istderrgv))->sv_flags & 0xff)) ==
SVt_PVGV)
&& ((0 + ((my_perl->Istderrgv))->sv_u.svu_gp)->gp_io)
&&
((XPVIO *) (((0 + ((my_perl->Istderrgv))->sv_u.svu_gp)->gp_io))->
sv_any)->
xio_ofp
? ((XPVIO
*) (((0 +
((my_perl->Istderrgv))->sv_u.svu_gp)->gp_io))->sv_any)->
xio_ofp : Perl_PerlIO_stderr (my_perl)), message, msglen);
Perl_my_exit (my_perl, 1);
}
do
{
(my_perl->Iin_eval) = (((cx)->cx_u.cx_blk.blku_u16) & 0x7F);
optype = (((cx)->cx_u.cx_blk.blku_u16) >> 7);
(my_perl->Ieval_root) = cx->cx_u.cx_blk.blk_u.blku_eval.old_eval_root;
if (cx->cx_u.cx_blk.blk_u.blku_eval.old_namesv)
Perl_sv_2mortal (my_perl, cx->cx_u.cx_blk.blk_u.blku_eval.old_namesv);
}
while (0);
namesv = cx->cx_u.cx_blk.blk_u.blku_eval.old_namesv;
oldcop = cx->cx_u.cx_blk.blku_oldcop;
restartjmpenv = cx->cx_u.cx_blk.blk_u.blku_eval.cur_top_env;
restartop = cx->cx_u.cx_blk.blk_u.blku_eval.retop;
if (gimme == 2)
*++newsp = &(my_perl->Isv_undef);
(my_perl->Istack_sp) = newsp;
do
{
if (((my_perl->Idebug) & 0x00000004))
Perl_deb (my_perl, "%s scope %ld (savestack=%ld) at %s:%d\n", "LEAVE",
(long) (my_perl->Iscopestack_ix), (long) (my_perl->Isavestack_ix),
"..\\pp_ctl.c", 1753);
Perl_pop_scope (my_perl);
}
while (0);
/* LEAVE could clobber PL_curcop (see save_re_context())
* XXX it might be better to find a way to avoid messing with
* PL_curcop in save_re_context() instead, but this is a more
* minimal fix --GSAR */
(my_perl->Icurcop) = oldcop;
if (optype == OP_REQUIRE)
{
(void) ((SV **)
Perl_hv_common_key_len (my_perl,
(((0 + ((my_perl->Iincgv))->sv_u.svu_gp)->gp_hv ? (0 +
((my_perl->Iincgv))->sv_u.svu_gp)->gp_hv : (0 + (Perl_gv_add_by_type (my_perl, ((my_perl->Iincgv)), SVt_PVHV))->sv_u.svu_gp)->gp_hv)), (((const char *) (0 + (namesv)->sv_u.svu_pv))), (((namesv)->sv_flags & 0x20000000) ? -(I32) ((XPV *) (namesv)->sv_any)->xpv_cur : (I32) ((XPV *) (namesv)->sv_any)->xpv_cur), (0x04 | 0x20), (&(my_perl->Isv_undef)), (0)));
/* note that unlike pp_entereval, pp_require isn't
* supposed to trap errors. So now that we've popped the
* EVAL that pp_require pushed, and processed the error
* message, rethrow the error */
Perl_croak (my_perl, "%" "-p" "Compilation failed in require",
((void *) (exceptsv ? exceptsv :
Perl_newSVpvn_flags (my_perl, ("" "Unknown error\n" ""),
(sizeof ("Unknown error\n") - 1),
0x00080000))));
}
if (in_eval & 4)
{
Perl_ck_warner (my_perl, (12), "\t(in cleanup) %" "-p", ((void *) (exceptsv)));
}
else
{
Perl_sv_setsv_flags (my_perl,
(*
((0 +
((my_perl->Ierrgv))->sv_u.svu_gp)->gp_sv ? &((0 +
((my_perl->
Ierrgv))->
sv_u.svu_gp)->
gp_sv) : &((0 +
(Perl_gv_add_by_type
(my_perl,
((my_perl->Ierrgv)), SVt_NULL))->sv_u.svu_gp)->gp_sv))), exceptsv, 2 | 512 | 1024);
}
(my_perl->Irestartjmpenv) = restartjmpenv;
(my_perl->Irestartop) = restartop;
do
{
if (((my_perl->Idebug) & 0x00000004))
{
int i = -1;
JMPENV *p = (my_perl->Itop_env);
while (p)
{
i++;
p = p->je_prev;
}
Perl_deb (my_perl, "JUMPENV_JUMP(%d) level=%d at %s:%d\n", (int) 3, i,
"..\\pp_ctl.c", 1783);
}
(void) 0;
if ((my_perl->Itop_env)->je_prev)
longjmp ((((my_perl->Itop_env)->je_buf)), (((3))));
if ((3) == 2)
PerlProc_exit((my_perl->Istatusvalue));
PerlIO_printf (Perl_PerlIO_stderr (my_perl), "panic: top_env, v=%d\n", (int) 3);
PerlProc_exit((1));
}
while (0);
/* NOTREACHED */
}
}
Perl_write_to_stderr (my_perl, exceptsv);
Perl_my_failure_exit (my_perl);
/* NOTREACHED */
} |
From @bulk88 |
From @bulk88forgot to add, the test program the above post was done with is "-e |
From @bulk88On Sun Apr 08 11:45:45 2012, bulk88 wrote:
# define ASSERT_CURPAD_ACTIVE(label) \ |
From @bulk88 |
From @bulk88On Sun Apr 08 12:30:34 2012, bulk88 wrote:
On my perl 5.12, curpad and comppad are non-null at the beginning ( At At I might research it more today, or not, IDK. If someone is else is |
From @bulk88On Sun Apr 08 13:17:36 2012, bulk88 wrote: In 5.12, curpad is added with the following callstack.
At this point, On entry to perl_clone_using On entry to pp_fork at This is now an opcode problem. _____________________________________________
|
From @cpansproutOn Sun Apr 08 13:17:36 2012, bulk88 wrote:
Thank you very much for researching this. I’m afraid I don’t have -- Father Chrysostomos |
From @bulk88On perl 5.12 PP(pp_entereval) opcode sets compcv. On perl 5.15 it doesn't. entereval calls, S_doeval. On 5.15, at, hmm, head git 5.15 doesn't match DEBUG_x(dump_eval()); In both 5.12 and 5.15, startop is null pointer on entry and remains |
From @bulk88Here is where in LEAVE_with_name compcv is set to null. I guess a NULL |
From @bulk88 |
From @bulk88On Sun Apr 08 15:12:29 2012, bulk88 wrote:
compcv is null on entry to S_doeval in 5.15 and 5.12. I will guess the bug is either a lack of a extra mortal/scope/save, or |
From @khwilliamsonOn 04/08/2012 04:28 PM, bulk 88 via RT wrote:
Thanks very much for your tracking this down. I'm hopeful that someone karl williamson |
From @rjbsThanks /very much/ for this detective work! It's really appreciated over here. -- |
From @cpansproutOn Sun Apr 08 15:28:23 2012, bulk88 wrote:
The change that caused this crash (see subject) intentionally restored
The purpose of the evalcomp scope (which I added earlier) is to undo Earlier, you wrote:
http://perl5.git.perl.org/perl.git/blob/03d9f026ae253e9e69212a3cf6f1944437e9f070:/sv.c#l13284
http://perl5.git.perl.org/perl.git/blob/03d9f026ae253e9e69212a3cf6f1944437e9f070:/pad.h#l339
That may be the problem. What happens if you apply the attached patch? This is a bit of a stab -- Father Chrysostomos |
From @cpansproutInline Patchdiff --git a/pad.h b/pad.h
index ddde4d7..aa6521f 100644
--- a/pad.h
+++ b/pad.h
@@ -338,7 +338,7 @@ Clone the state variables associated with running and compiling pads.
* sub's CV or padlist. */
#define PAD_CLONE_VARS(proto_perl, param) \
- PL_comppad = MUTABLE_AV(ptr_table_fetch(PL_ptr_table, proto_perl->Icomppad)); \
+ PL_comppad = av_dup(proto_perl->Icomppad, param); \
PL_curpad = PL_comppad ? AvARRAY(PL_comppad) : NULL; \
PL_comppad_name = av_dup(proto_perl->Icomppad_name, param); \
PL_comppad_name_fill = proto_perl->Icomppad_name_fill; \ |
From @janduboisOn Mon, 09 Apr 2012, Father Chrysostomos via RT wrote:
Excellent, this does indeed fix the crashes! Cheers, |
From @cpansproutOn Mon Apr 09 19:59:06 2012, jdb wrote:
I’ve pushed it to the smoke-me/109718 branch. Let’s see what the Believe or not, I’ve not even tried compiling it yet. :-) -- Father Chrysostomos |
From @cpansproutOn Mon Apr 09 20:38:49 2012, sprout wrote:
Could this getppid.t failure have anything to do with it, or is it On Apr 9, 2012, at 9:53 PM, George Greer wrote:
-- Father Chrysostomos |
From @bulk88On Mon Apr 09 18:18:56 2012, sprout wrote:
http://perl5.git.perl.org/perl.git/blob/03d9f026ae253e9e69212a3cf6f1944437e9f070:/sv.c#l13284
http://perl5.git.perl.org/perl.git/blob/03d9f026ae253e9e69212a3cf6f1944437e9f070:/pad.h#l339
I haven't tested your fix. compad becomes the pad in win32_start_child. Perl_set_context(my_perl); ((my_perl->Isys_intern).pseudo_id) = GetCurrentThreadId(); /* push a zero on the stack (we are the child) */ /* continue from next op */ The reason why compad is null/not found in pointer table is because In 5.12, the line that triggered the cloning is PERL_ARGS_ASSERT_SV_DUP; if (!sstr) If compcv is null, sv_dup does nothing, so comppad wont be added to the I will try to run the patch sometime in the next day or 2 and i'll give |
From @iabynOn Mon, Apr 09, 2012 at 06:18:57PM -0700, Father Chrysostomos via RT wrote:
I don't think that's the most correct fix. The intention is that whatever I suspect the correct fix is to ensure the current CV is always cloned; Do you want to run with this, or should I? -- |
From @cpansproutOn Tue Apr 10 04:30:45 2012, davem wrote:
If you are offering to take over, please go ahead. :-) I need to ask the Windows users: Does perl -efork crash? If it doesn’t, can you (Dave Mitchell) explain why not? -- Father Chrysostomos |
From @cpansproutOn Tue Apr 10 03:55:30 2012, bulk88 wrote:
I understood that. I realise I may have been a bit terse.
http://perl5.git.perl.org/perl.git/blob/03d9f026ae253e9e69212a3cf6f1944437e9f070:/sv.c#l13282
http://perl5.git.perl.org/perl.git/blob/e42956688f2e0df936f1a42811962946e4e185bf:/sv.c#l11329
What I was trying to say is that PL_compcv is *supposed* to be null in The code for setting PL_comppad in the child thread was making untenable -- Father Chrysostomos |
From @greergaOn Mon, 9 Apr 2012, Father Chrysostomos via RT wrote:
Probably unrelated as I've seen it before. It's like the tie deepcopy -- |
From @iabynOn Tue, Apr 10, 2012 at 06:04:36AM -0700, Father Chrysostomos via RT wrote:
Whoops, that was silly of me :-)
Because then the currently executing sub is PL_main_cv which is -- |
From @iabynOn Tue, Apr 10, 2012 at 12:21:30PM +0100, Dave Mitchell wrote:
Actually, I misread a bit of code in cx_dup(); the CV *is* duplicated in -- |
From @bulk88The av_dup patch worked for me. The getppid test isn't done on windows. |
From @cpansproutOn Tue Apr 10 07:08:11 2012, davem wrote:
I’ve applied it as 253649d. -- Father Chrysostomos |
|
@cpansprout - Status changed from 'open' to 'resolved' |
From @cpansproutThis ticket might not be completely resolved yet. In <612945.556640625-sendEmail@hpwin7>, Steve Hay’s smoke report says: Failures: (common-args) -DCCTYPE=MSVC100 Steve, are you able to find out when this particular failure started? I -- Father Chrysostomos |
From @steve-m-hayOn 11 April 2012 22:11, Father Chrysostomos via RT
Not easily. I've only recently restarted smoking again after a long I do notice that smokes from me before the one above were failing In the absence of much recent smoke history on this machine, I have |
From @kmxFYI: here is how goes ./t/op/fork.t on Windows 7 + gcc-4.4.7/32bit - tested op/fork.t ......................................................... 23/? PROG: No crash, just the one failure (which AFAIK exists some time and is related -- |
Migrated from rt.perl.org#109718 (status was 'resolved')
Searchable as RT109718$
The text was updated successfully, but these errors were encountered: