Regression between 9th and 19th december (game Beyond Good and Evil)

[fafner2000]

Hello,

Short story:

Something broke the GOG version of the game Beyond Good and Evil [1] between commits 1c282c2 (2025-12-09) and f2660cd (2025-12-19).

I will probably try to find the exact commit where the regression bug is, but I'll have to find some time to do that. However, if in the meantime you have a possible fix, just tell me and I'll make sure to test it.

Long story:

Last 10th december, I installed the GOG version of the game Beyond Good and Evil [1] on a Debian 13 (Trixie) Raspberry PI. At this point the game worked fine with some light performances issues.

I uninstalled both wine and box64 to test hangover. It failed miserably, so I just cleaned up the mess and reinstalled box64 and wine. Total failure: the game still starts and can load games, but there was heavy graphical glitches and often camera positioning issues, as well as unexplained sudden deaths. I finally traced the problem to box64, as uninstalling box64 and reinstalling the version downloaded the 10th right from the system cache fixed everything. Luckily, it hadn't been purged yet; you can be sure that I swiftly made a backup.

I used the application Pi-Apps which automates everything so I don't have to spent hours finding the magic incantations to make things work. However, for some reason, it installs the latest commit available, probably not caring that it might not be an actual release. Given the names of the .deb files involved, I deduced that the regression bug was somewhere between commits 1c282c2 (2025-12-09) and f2660cd (2025-12-19). I made a last try today with commit 050e0f9 (2025-12-28) and it still has the same issues.

Note: The game is 32 bits, but the wine version installed on the Raspberry PI is 100% 64 bits and uses some kind of 32 bits emulation similar to what Microsoft Windows does.

[1] https://www.gog.com/en/game/beyond_good_and_evil

[ptitSeb]

I don't own this game, so cannot test. Have you tested the game with latest box64? is it still broken? Because the current version seems very stable for me now.

You might need to build box64 from source to really bisect the issue.

(note: this is using Wine+new WOW64 indeed, to stay in 100% 64bit lands, similar to MS Windows)

[fafner2000]

Update: I compiled with commit 09bdbcb (latest) and 050e0f9 (the one where I observed a problem), and the problem has disappeared from both. However, I compiled with default options, and the performances are abyssal (less that 1 FPS) and even exiting the game takes something like 1 minute. I'll try to find how to compile with release options.

Edit 1: I suspect that the missing option is -D ARM_DYNAREC=ON
Edit 2: I finally compiled with options "-D RPI5ARM64=1 -D CMAKE_BUILD_TYPE=RelWithDebInfo" (which seems to imply dynarec) and the problem has reappeared with commit 050e0f9.

[ptitSeb]

Update: I compiled with commit 09bdbcb (latest) and 050e0f9 (the one where I observed a problem), and the problem has disappeared from both. However, I compiled with default options, and the performances are abyssal (less that 1 FPS) and even exiting the game takes something like 1 minute. I'll try to find how to compile with release options.

Edit 1: I suspect that the missing option is -D ARM_DYNAREC=ON Edit 2: I finally compiled with options "-D RPI5ARM64=1 -D CMAKE_BUILD_TYPE=RelWithDebInfo" (which seems to imply dynarec) and the problem has reappeared with commit 050e0f9.

I guess I have compiled without Dynarec!
Yeah, see the edit, you found the right parameters.

Now, the bug in on 050e0f9 ok, but is it solved with 09bdbcb ?

[fafner2000]

I am compiling this commit right now, it takes forever :-D

[fafner2000]

Bug still present on commit 09bdbcb.

[ptitSeb]

Ah ok. I might buy the game on GoG (it's on sale) and debug on my side. Probably an issue on x87 emulation, it's tricky to debug...

[fafner2000]

Bug not present on commit 1c282c2.

Edit 1: Tomorrow afternoon I'll have some time to bisect more precisely.
Edit 2: Present on b9a9317.
Edit 3: Present on 3b9edf0.
Edit 4: Present on 57466cf.

[ptitSeb]

Culprit is eb743a1 wich is a complex commit. I'll work on that tomorrow.

[ptitSeb]

I have a fix. I'll do some more testing tomorrow and will push it.

[ptitSeb]

Ok, it's fixed for me. You can update box64 and try on your side.

[g7ruh]

Thanks for the update, it fixes a problem in PlanePlotter where the graphics is messed up. Checked this morning for any reported issues and arrived here.

The problem:

which displays correctly after compiling the recent update:

[ptitSeb]

Thanks for reporting this @g7ruh :)

[fafner2000]

Fixed on my side too.

I have other games that have subtle (or sometimes no so subtle) bugs, is there some built-in utility that I can use to attempt to examine how box64 recompile them ?