httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "William A. Rowe, Jr." <wr...@rowe-clan.net>
Subject [PATCH] Cleaning mod_cgi.c
Date Tue, 11 Sep 2001 04:16:21 GMT
Since Brane has picked up on my efforts to cleanup the apr_proc_create fooness,
I'd like to push to clean this out sooner rather than later.

The ap_cgi_build_command optional function would be implemented on platforms 
that don't like the 'unix' way of building the command and argv[] for cgi 
applications.  So on my usual platform, it would be registered by mod_win32.c,
which I'm in the process of coding.

This allows us (on win32) to rip all of that ScriptInterpreterSource cruft from
the core, and deposit it on ourself.  It allows me to build a proper argv[],
from the registry, even if the association is as odd as 
  "%SystemRoot%\cscript.exe "%1" %* //nologo" which is how bad things can get 
around here ;)

The default_build_command code I dropped in should cover the unix (and suexec)
case, but there were some odd discrepancies amoungst all the code repetition.
It folds the old, seperate command and argv processing functions into one,
which they should have been in the first place (who's on first ;-?)  So I'm
asking for anyone who enjoys playing with suexec and SSI to try this patch and
whack on it a bit.

If reports come up good, the code for win32 should be ready by Thursday night,
I hope.  I'll drop it in all at once.  ITMT, coders hacking with Win32's CGI
execution (Brane?) are forewarned that this new code is in the pipe.  That
doesn't mean I won't encourage anyone to keep hacking at apr_proc_create for
Win32 - especially if you care to implement the 16 bit process downgrade on
failure.  I have to find a clever way to get a 32/16 bit console thunk into 
the loop, and I half expect that to go right into mod_win32.c.

Bill


Mime
View raw message