Florian Weimer <fwei...@redhat.com> writes: > * Sam James: > >>> Has anyone performed experiments to determine the impact of this change >>> on typical free software distributions? >> >> I filed https://gcc.gnu.org/PR117298 for an issue Joseph noticed in one >> of the GCC tests (that is actually an improvement, but a missed opt for >> older standards). I haven't done any sort of testing but am curious >> about it as well. >> >> I could do such a test for code size en-masse (and perhaps maybe even >> check where the image changed at all). Runtime performance is far harder >> for me to do at scale though. We can use significant code size changes >> as a proxy for interesting candidates to investigate though. >> >> What are you thinking of? > > Mostly compatibility with older configure scripts. There are the new > keywords bool/true/false, and the -Werror=deprecated-non-prototype > default that could alter configure test outcomes. Maybe there's > something else I'm missing? > > Some estimate of the build failure would also be helpful. Is it the > same as, e.g. the GCC 12 to GCC 13 update, or is it igher?
It's pretty large so far. Between 12 and 13, the main issues were: libstdc++ transitive include changes and a small number of -Werror breakages. See https://bugs.gentoo.org/show_bug.cgi?id=880545 and https://bugs.gentoo.org/showdependencytree.cgi?id=880545&hide_resolved=1 (*). I think there'll be a lot more, the testers are currently blocked on some core packages not building. I've done no configure test checking yet, but I have seen a handful (3, I think) of packages which failed to build that indicate a configure test went wrong. (*) That tracker was started initially for when Clang started warning about it and so on and didn't have that many blockers until I revived it when the GCC default changed the other week. thanks, sam
