> For Apache 2.0.x these functions return 255, the default value in 2.2. > Setting > Apache2::CmdParms::override_opts dies for 2.0 and the override_opts parameter > to add_config is simply ignored in 2.0. > > When adding tests I stumbled. Where do I have to add the -defines trick in > this case?
well, I'm not sure I follow you, then. the only real reason for wanting what you asked for in the first place was to prevent httpd from crashing on startup because of configurations options it doesn't understand. if you can start httpd then you should be able to vary what you test against using have_min_apache_version() in the test itself. and it looks to me like you can test what you're after without needing to start httpd with configurations that differ between 2.0 and 2.2, unless I'm completely misunderstanding things. so, then, why do you need a config that causes 2.0 not to start as the only way you can test your new functionality. that's the real question :) > > Btw, does anyone know why $r->add_config always passes a path in cmd_parms? I > have tried passing also NULL pointers. For some configuration directives that > causes segfaults but sometimes it's useful (for example if a <Location> block > is to be passed to $r->add_config). I'd like to extend the interface to pass > a NULL pointer in cmd_parms if an empty string is passed to add_config. philippe is the man here. > > Would such a patch have any chances to get into mod_perl? well documented and understood patches always stand a good chance, provided they don't break back compat. --Geoff
