Comment 23 for bug 988819

Revision history for this message
Steve Langasek (vorlon) wrote :

I have serious misgivings about the particular approach taken here, in SRU or otherwise. Why does mod-security.load need to hard-code libxml2.so.2 *at all*? Isn't that why DSOs are dynamically linked to their dependent libraries (which mod_security2.so is)? And why is libxml2.so.2 being special-cased here, when mod_security2.so is also linked against liblua5.1.so.0? Did anyone try to fix this by just removing the LoadFile line completely? That would be a much safer fix than SRUing the core of apache2, and would probably fix the problem that this SRU seems to be stalled waiting for indeterminate regression testing of "corner cases".

A naive test here of commenting out this 'LoadFile' line shows that the apache config validates, and apache2-mpm-prefork starts without any problems.