On Wed, May 10, 2023 at 09:49:25PM -0500, Linus Torvalds wrote: > On Wed, May 10, 2023 at 9:01 PM Dave Chinner <david@xxxxxxxxxxxxx> wrote: > > > > [ and now with the correct cc's. DOH! ] > > [ Bah. And now with the correct cc's on the reply too. ] > > We do not consider build warnings even remotely "harmless". > > That sounds like you will fail all build farms that happen to have > this compiler version. ...and which version is that? The build robot report just says ia64 without specifying any details about what compiler was running, etc: https://lore.kernel.org/linux-xfs/20230510165934.5Zgh4%25lkp@xxxxxxxxx/T/#u I'll see if I can fix it, but there's not enough information to reproduce the exact circumstances of the warning. I've been building this branch with gcc 11.3 on x86_64, aarch64, ppc64le, s390x, and riscv64, and none of them complained. I normally build the fs/xfs code with most of the W=12e options enabled, though I confess to turning off the warnings that trigger on random bits of source code outside xfs. I don't have access to any ia64 environment, let alone whatever build farm Intel has. > Which in turn just means that I'd have to revert the commit. The PR message didn't specify what warning was being ignored. Assuming it's the one the build robot found, would you actually revert a commit over a build warning about unreachable code *only* on ia64? > The fact that you seem to think that build warnings don't matter and > are harmless AND you imply that you can just leave some compiler > warning to the next release just makes me go "No, I don't want to pull > this". > > So I pulled this. It built fine for me. But then I went "Dave says he > isn't even bothering to fix some build warning he seems to know about, > and isn't even planning on fixing it until 6.5, so no, I think I'll > unpull again". The real fixes for the code that I turned off are sufficiently involved that we had to have a session about it at LSFMM. There's no way to merge it now; the window is closed. --D > When you decide that compiler warnings matter, please re-send. > > Linus
