On Sun, 2002-05-26 at 13:18, Brian Pane wrote: > On Sat, 2002-05-25 at 20:17, Colm MacC�rthaigh wrote:
> > patch 1 (mod_cgid.c) fixes 7810/9039/mod_cgid, it just works. > > Patch 1 didn't work when I tried it. The problem was that it depends > on mod_userdir's get_suexec_id_doer() to create a proper suexec > config based on the userdir. But mod_userdir's get_suexec_id_doer() > function never is invoked, because mod_suexec's get_suexec_id_doer() > is called first. > > I modified mod_userdir to insert its suexec identity function in > front of mod_suexec's. The suexec+cgid+userdir combination now works > properly in my test environment, but I'd like a second opinion before > I commit. Actually, as Colm pointed out in PR 9083, the precedence of mod_userdir vs mod_suexec is a more general design issue, independent of mod_cgid. With patch1 and without a mod_userdir change, the combination of cgid+userdir+suexec works as long as there's no global SuexecUserGroup directive. That brings cgid up to par with cgi, so I'll commit the cgid patch. --Brian
