Re: [Patch]: Get mod_cgid working.

[rb...@covalent.net]

On Fri, 26 May 2000, Paul J. Reder wrote:

> The attached patch fixes a few small mod_cgid problems. Most are
> just macro changes. A couple are related to changes function parms.
> Moving the ap_destroy_pool call fixes a core dump on the second cgi
> to be run (the pool was being destroyed too soon).
> 
> Preliminary test results show doubled performance when using mod_cgid 
> over mod_cgi on Linux. I will be running a larger set of tests with
> a variety of parms to get a better idea of the results.

Not that I'm disputing the double performance, but I would really like to
understand why that is.  The original reasoning behind mod_cgid was that
threaded processes when forked have a nasty habit of creating all of the
threads in the new process and then killing them off one-by-one.  I find
it hard to believe that Linux was falling into this behavior (only because
Linux threads and processes are the same).

I think a better test (potentially) would be to modify mod_cgi to use
fork1 or fastfork on platforms that support them (AIX and Solaris can't
remember which does which) and re-compare.  Those two routines basically
make sure not to create the threads in the new process.

I'll be looking the patch over very soon.

Ryan