Yavor Doganov <ya...@gnu.org> (23/12/2009): > Don't bother, please -- I managed to reproduce the ICE on fencepost > with a manually built GCC 4.4.2. Thanks for your efforts.
Okay. And you're welcome, again. ;) > Now, what are the options to deal with this FTBFS until the GCC bug > is fixed? > > 1) Compile with -O0 on amd64/kfreebsd-amd64. This would be a Debian > Policy violation; I'm not sure if it is allowed in cases like this. I would do that. Working around compiler bugs until they're fixed is AFAICT very OK. Just make sure it's properly documented in e.g. changelog and debian/rules, so that arch-specific stuff can be cleaned up afterwards without having to wonder why that was put into place in the first place. Mentioning the compiler bug number would be nice, for example. > 2) Build-depend on gobjc-4.3 on these architectures and compile the > package with that version (assuming, of course, that the ICE doesn't > happen with 4.3). This is probably slightly clumsy, and would > prevent the eventual removal of gcc-4.3 in the foreseeable future. You already have the bottom line: not really a solution (unless agreed with the release team/compiler folks that a whole bunch of packages need such a different version, for whatever reasons). Mraw, KiBi.
signature.asc
Description: Digital signature