Greetings.
I am hitting a complex Registry application with a number of almost
simultaneous, long running requests
(WinNT SP6 - VC++6 SP3 - perl5.8.0 RC1 - latest dev snapshot for modperl -
httpd-2.0_20020530041218.tar.gz for apache - everything built debug - a few
date.t tests fail on nmake test).
The crash happens in perllib.c at:
EXTERN_C PerlInterpreter *
perl_clone_host(PerlInterpreter* proto_perl, UV flags) {
dTHX;
CPerlHost *h;
h = new CPerlHost(*(CPerlHost*)PL_sys_intern.internal_host); //<==
BLAAAMM!
proto_perl = perl_clone_using(proto_perl, flags,
h->m_pHostperlMem,
h->m_pHostperlMemShared,
h->m_pHostperlMemParse,
h->m_pHostperlEnv,
h->m_pHostperlStdIO,
h->m_pHostperlLIO,
h->m_pHostperlDir,
h->m_pHostperlSock,
h->m_pHostperlProc
);
[...]
Coming from modperl_interp_new.
I have not been able to reproduce this on simpler applications, so far (but
I will try).
I can also crash the same app under normal usage, however, but not
reproducibly so: a constant factor in the crashes (not in the one described
above, though) appears to be the appearance of "bad header" warnigns in
error.log: the crash follows shortly.
Note that the same app runs fine (no warnings whatsoever under modperl1).
I suspect that the threading code is to blame on all these crash counts: my
app has 3 frames, all hitting the same URL in rapid succession and this
appears to be brushing modperl the wrong way.
All of the above happens with the default intepreter pool settings: if I try
to
tune it, the application goes southwest immediately, only misteriously so -
it just spins, without warnings or anything.
Cheers,
alf
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
