httpd

I'm not going to be all high and mighty then say that tests should have caught this since Apache has tests. I am not sure why RewriteCond expr is even valid at all since expr typically needs an expression to come after it like RewriteCond expr "'cache/%{md5:%{REQUEST_URI}?%{QUERY_STRING}}.html' =~ /(.+)/". The bug appeared to be introduced on July 7 with a backport from trunk: https://github.com/apache/httpd/blame/ed99ef021de902363c36af... It took me a second but the fix addresses the case where rc==0. If statement is less than 0. Therefore, rc==0 should indicate no match. https://github.com/apache/httpd/commit/8abb3d06b23975705ebcf... I suppose because there is also the conditional for err?

Based on https://github.com/apache/httpd/commits?author=covener, either the committer never writes tests, or this project just doesn't do testing at all. Nothing here would pass a code review at my company. Totally insane.

It seems like the bug was originally introduced here: https://github.com/apache/httpd/commit/dd98030cb399e962aa605... This is seemingly a well-intentioned cleanup that misunderstood the branching logic. The original code was normalizing `rc < 0` to COND_RC_NOMATCH (0), but leaving `rc >= 0` as the original value. However the new code accidentally normalizes `rc >= 0` to COND_RC_MATCH (1) while it should be normalizing `rc > 0` to COND_RC_MATCH (leaving 0 as COND_RC_NOMATCH). There are a few other logic changes in this patch which should likely be reviewed carefully. For example https://github.com/apache/httpd/commit/dd98030cb399e962aa605... does `rc <= COND_RC_MATCH` which matches both match and no match. Presumably it is checking for COND_RC_STATUS_SET but it seems like and odd way to write `rc == COND_RC_STATUS_SET`. Maybe the intention