Re: Initializer

[Doug MacEachern <do...@opengroup.org>]

Dirk-Willem van Gulik <di...@elect6.jrc.it> wrote:

> 
> > It is also a bit tricky that it is the second call to the function that is
> > the real call.  I find that I have to ignore the first call and just take
> > the second for things to work correctly.
> > 
> > I recall that someone fixed this *feature* at some point resulting in
> > mod_perl breaking.
> 
> I tried to fix it, and it indeed stoped mod_perl, fastcgi, include and the
> oracle sql module from working :-( the first two for not exactly trivial
> reasons. This could turn out to be an expensive legacy problem.

mod_perl should be okay now.  During the 1.3b1-dev cycle, when apache
was switched to one init_modules() call at startup, that broke <Perl>
sections.  I complained, and we agreed to go back to calling
read_config/init_modules twice, not just for mod_perl though.
During that, I re-worked things so <Perl> sections can survive either 
way.  
mod_perl still avoids the second init_modules() at startup, mostly so
we don't have to wait for all that pre-loaded Perl code to be
re-parsed/compiled.  For the same reason, by default mod_perl's
init_modules() is a noop during restart, unless httpd.conf has 
'PerlFreshRestart On'.  In which case, the pre-loaded code will be
re-compiled. 
 
-Doug