From: Neal Becker (ndbecker2_at_gmail.com)
Date: Mon Dec 22 2008 - 11:55:16 PST
On Monday 22 December 2008, Paul H. Hargrove wrote: > With all due respect to the person who raised the question: a link failure > is NOT any indication that such buffer overflows are present. It is an > indication of a toolchain that creates implicit dependencies and broke > linking of many shared libraries and dynamically loadable modules other > than BLCR. > > Neal, have you or anyone else tried removing -fno-stack-protector to see if > things work correctly on Fedora 9 or 10? Its possible they might. > > If somebody can point to the proper way to resolve the linkage problem, > without also breaking things for platforms w/o stack-protector support in > gcc, then I'd be happy to apply the fix. > I could try it, but how does it break? I'm asking because if it only breaks on some arch, does it break in compile? Or only runtime? If the former, I could easily try to build it using the compile farms already available. If it only breaks in testing, I'd have to do some more.