On Apr 29, 2014, at 7:41 AM, Jan Kaluža <[email protected]> wrote: > > That's what we do with current patch I think, don't we? In the patch, we > create "char *match_name" which is NULL when the worker_name is not regex and > contains the escaped name if regex is used (with "$N" replaced by '*'). > > ap_proxy_get_worker() later checks the match_name and if it's not NULL, it > tries regex matching using ap_proxy_strcmp_ematch(). > > The actual problem is that ap_proxy_get_worker() cannot find out proper > proxy_worker, because this method is not regex aware. To make it regex aware, > we have to distinguish regex workers during creation,
right, hence the flag. > replace the regex variables ($N -> *) in some safe manner and allow > ap_proxy_get_worker() to find the right worker using this new info. And > that's what the patch does. Why replace? Why can't we store the name as is? The version I looked at had some 'use_regex' logic outside of ap_proxy_get_worker(), right before we call it, in fact. Maybe I'm seeing an older version of the patch?
