Martin Husemann <martin%duskware.de@localhost> writes: > Sorry folks, I broke the build big time - and this time it was kinda on > purpose. > > In order to systematically track down all changes needed to fix PR 49074 > (but for all architectures, and future proof), I added a link time test > exercising the C++ 2011 <atomic> operations. > > The list of architectures this did NOT break on first try was: amd64, ppc64 > and sparc64 (guess which two I tried upfront?). > > This should now be slightly better (i386 should pass as well, probably quite > a few others), but it will take a few days to track down and squash the > stranglers. It doesn't really bother me that the build is broken for me compiling, but it's problematic to have other people committing things during an extended period of breakage, because the breakages caused by that are then harder to isolate (and breakage seems to tend to be more numerous). So perhaps you could disable the new <atomic> stuff in what's checked in, and then re-add it locally while you work through builds on various platforms. That would seem to be about as much work for you, but not impose breakage on others, and more importantly not cause the secondary breakage. (Actually, it looks like the i386 build was already broken from other things when you added this.)
Attachment:
pgpF9t7VVu_Ij.pgp
Description: PGP signature