http://bugs.winehq.org/show_bug.cgi?id=57986
Chiitoo chiitoo@gentoo.org changed:
What |Removed |Added ---------------------------------------------------------------------------- Fixed by SHA1|b52ff2d69961bd1f3adc7ba386f |659e3b3e14e45526109870bf0da |179b12af26481 |70d63b10e5641 Resolution|--- |FIXED Status|UNCONFIRMED |RESOLVED
--- Comment #4 from Chiitoo chiitoo@gentoo.org --- It seems some confusion here was caused by different behaviour between builds using different GCC versions (and, hopefully, 'ccache').
While it seemed as if the issue was "moving goal posts", I never saw different behaviour on re-running the application on a single same build. After a while I realised I was seeing different behaviour with an older build of Wine release 10.4, which had been built using GCC 14.2.1.
After even more bisecting back and forth, and still feeling like something wasn't right with the "goal posts", I started suspecting 'ccache' having an effect as well, which I had meant to disable already a while ago for this (not that it even helps with these CPU/HDD speeds here (building in RAM in addition still seems to give me some gains though, depending on changes of course).
I can still see differences between the GCC versions though, so there might be something to investigate on the GCC side as well, but what comes to the Wine part, this issue seems to have been resolved with GCC 15.1.0 at Wine commit 9141dd4d850 [4] (included in release 10.7), and for current GCC 14.2.1_p20250419 (as well as 13.3.1_p20250418) a bit later at Wine commit 659e3b3e14e [5]. Things improved or changed for 14.2.1 earlier at some point already too, but the initial state for the window was still incorrect (looking at the later change, I am quite unsure why 15.1.0 does better without it... but maybe that is for another investigation time).
Not sure if it's possible for GCC to use objects from a cache that was populated by a build with a different version of GCC, though I suppose that could explain some of the strange changes in behaviour I was seeing when switching the version and moving between commits again.
I also went back to the regression SHA, to make sure the issue starts with both versions the same, and that does seem to be the case (also tested some with GCC 13.3.1_p20250418).
In any case, right now at current git master, 09539a3af5f [6], the window behaves as expected regardless of which of the 3 GCC version I build Wine with.
Thank you!
4. https://gitlab.winehq.org/wine/wine/-/commit/9141dd4d850210788dcc63838abeca2... 5. https://gitlab.winehq.org/wine/wine/-/commit/659e3b3e14e45526109870bf0da70d6... 6. https://gitlab.winehq.org/wine/wine/-/commit/09539a3af5f806bf319917cafff2b07...