After several days of debugging I think I figured out that perl 5.8.3 with
a default config doesn’t work properly on Solaris 10.
I installed on a Solaris 10 (doing a major upgrade from a working Solaris
9 similar setup ) the following:
Apache/1.3.34 (Unix) PHP/5.1.2 mod_perl/1.29 using the Solaris malloc
(perl 5.8.3 configured with -Uusemyalloc (means: don’t use perl’s malloc)
and the rest default)
(All compiled with the Sunfreeware gcc 3.3.2 package and APACI for PHP and
RT 3.4.5 with RTFM-2.2.0RC2.
From time to time one of the Apache processes simply bails out with the
following message: “Out of memory!” (This message is perl related)
(grep through the docs,… seems that it’s perlio related (Changes5.8))
(seems to be after PHP activity in the same Apache)
I also tried with perl 5.8.8 (plus mod_perl and Apache rebuild ofcourse,…
no go ,… even more errors.
Tried running Apache with debugging perl in standalone,…
Tried perl profiling,…
No answer found,… Apache processes keep bailing out…
So I added the 2 option in the perl config to NOT use perlio (non-default)
and NO faststdio (non-default)
(Watch out here,… the perl config will use a different naming scheme for
it’s @INC dir’s. Solved this by editing the perl’s config.sh file before
the perl build. It also means that this will break any statically linked
Perl Modules from a same versioned Perl with different config options,
because the PerLio is gone)
As far I can see, repairing the RT required modules involves:
CPAN break’s on Term::Readkey (install it manually)
then use CPAN shell to force install (in this order):
Seems that the “Out of Memory!” error is still there !!
Arrghh,… I’m at a loss here,…
The Solaris installed perl isn’t an option because we need a more recent
one than Sun’s 5.8.4 (see above ,… 5.8.8 default fails to)
Any hints ? A working recipe ?