On Thu, 9 Oct 2025 at 17:12, Hans-Peter Nilsson <[email protected]> wrote: > > > From: Hans-Peter Nilsson <[email protected]> > > Content-Type: text/plain; charset="iso-8859-1" > > For transparency and closure: > > > I just don't get how those remaining > > config/cpu/*/atomicity.h files can be reached, considering > > that config/cpu/cris/atomicity.h was not compiled for no > > obvious reason. But, I also don't see the latter file > > wasn't compiled. I'll have to check myself, I can't let go > > of this. > > A-ha, you *added* a use in r16-3810-g6456da6bab8a2c; the > "regressing" commit! My POC-proving-not-being-compiled > patch in PR122172 was checked against r16-3809 when it was > indeed not compiled.
Yes, the change that caused cris-elf to start using config/cpu/generic/atomic_builtins was earlier: r16-427-g86627faec10da5 For some targets (including cris-elf), r16-427 meant that the value of atomicity_dir set by configure.host was overridden, and replaced with atomicity_dir=config/cpu/generic/atomicity_builtins. But you didn't notice any problems until the later r16-3810 where I introduced the code that triggers the -Wignored-attributes warnings. > The commit-date for r16-3810-g6456da6bab8a2c seems to be > "Thu Jul 17 22:02:45 2025 +0100", unexpectedly much earlier > than 2025-09-12 when my autester started screaming. Not > wrong, but made me think I fumble-fingered and was looking > at the wrong revision. Ah, that's the *AuthorDate*: commit 6456da6bab8a2c43e7899afda991589065d96595 Author: Jonathan Wakely <[email protected]> AuthorDate: Thu Jul 17 22:02:45 2025 Commit: Jonathan Wakely <[email protected]> CommitDate: Thu Sep 11 14:43:55 2025 I wrote (or finished writing) the patch in July but didn't push it until September. That happens with a lot of my patches, because I sit on them or forget about them in some local branch, then push when I think they're sufficiently baked/tested. My ~/.gitconfig contains: [format] pretty = fuller so that my git log output always shows both dates.
