On Thu, 22 Apr 2004, Stas Bekman wrote:
> stas@xxxxxxxxxx wrote:
> [...]
> > Index: extra.conf.in
> > ===================================================================
> > RCS file: /home/cvs/modperl-2.0/t/conf/extra.conf.in,v
> > retrieving revision 1.9
> > retrieving revision 1.10
> > diff -u -u -r1.9 -r1.10
> > --- extra.conf.in 15 Nov 2003 19:36:00 -0000 1.9
> > +++ extra.conf.in 22 Apr 2004 22:30:56 -0000 1.10
> > @@ -1,3 +1,33 @@
> > +# The following tests require more then 1 interpreters during the same
> > +# request:
> > +#
> > +# perls Test
> > +# -----------
> > +# 2 t/modules/apache_status
> > +# 2 t/filter/both_str_req_proxy
> > +# 2 t/modules/proxy
> > +#
> > +# the following tests will clone a new perl interpreter via
> > +# ithreads.pm regardless of how many interpreters mod_perl has:
> > +#
> > +# t/perl/ithreads
> > +# t/perl/ithreads2
> > +#
> > +# therefore we need at most 2 interpreters, the missing one loading on
> > +# demand, but we start only with 1, so the startup is quick,
> > +# especially since we immediately restart. We also want MaxSpare to be
> > +# the same as Max, since we have more than one test that requires more
> > +# than one interpreter, so don't waste time to kill and start a new
> > +# one later -- keep it around once spawned
> > +#
> > +# Adjust PerlInterpMax and PerlInterpMaxSpare if the requirements change
> > +<IfDefine PERL_USEITHREADS>
> > + PerlInterpStart 1
> > + PerlInterpMax 2
> > + PerlInterpMinSpare 1
> > + PerlInterpMaxSpare 2
> > +</IfDefine>
>
> Please test that it still works on windows, I tested only
> with prefork/worker. It should be the same, but may be
> there are special demands on win32 that I'm not aware of.
> Notice that it's OK when it'll be slow a bit when
> encountering one of the above listed 3 tested for the
> first time - it'll spawn a second enterpreter. This is
> actually a good change, since besides faster startups for
> threaded mpms (about 3 times faster on my machine with
> worker mpm), it also tests the startup of extra
> interpreters at run time (which as we already know is
> dreadfully slow).
Hi Stas,
This checks out fine on Win32 - thanks.
--
best regards,
randy
|
