wineserver (from Wine 8.0.1 i386) crashes with SIGSEGV (NULL pointer access) on Crostini (Debian Bookworm ARM Mediatek) using Box86 v0.3.9

[ikumayuki]

OS: ChromeOS Crostini (Debian 12 Bookworm)
CPU: Mediatek ARM
Box86 version: v0.3.9 fa59e74 built on Feb 13 2025 06:33:33 (From pi-apps repository)
Wine version: 8.0.1 (i386/x86)
Wine installation method: Manually installed from WineHQ Debian Bookworm .deb files (wine-stable-i386_8.0.1bookworm-1_i386.deb and wine-stable_8.0.1bookworm-1_i386.deb). Extracted and combined contents into ~/wine.}

I'm trying to run .exe on Chromebook by using box86 and wine.

wineserver consistently crashes with a SIGSEGV (segmentation fault) due to a NULL pointer access, immediately after being launched. This occurs both when wineserver is invoked by the wine command (e.g., during wine wineboot) and when wineserver is launched directly via Box86. This crash prevents Wine from initializing correctly and running any Windows applications. A "Double SIGSEGV" message is also observed in the strace output when wineserver crashes.

Logs/strace: https://paste.ee/p/ktVmDmQW

[ptitSeb]

Maybe it would be easier to not use box86 but use Box64+Box32 instead? That would remove the need to use armhf multilib complexities.

[ikumayuki]

Hi @ptitSeb,
Thank you again for your previous advice regarding using Box64+Box32 to avoid armhf complexities. I apologize if my initial reports were too lengthy.
I have since:
Cleaned my Crostini (Debian Bookworm on Mediatek ARM) environment.
Cloned the Box64 repository, checked out the commit that seems to correspond to v0.3.5 (specifically b1381125 as reported by box64 --version after build and install, built on May 20 2025 23:51:00). I built it with cmake .. -DARM_DYNAREC=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DBOX32=ON.
Manually installed Wine 9.0.0 (stable i386 from WineHQ .deb files for Debian Bookworm) into ~/wine.
Created a wrapper script at /usr/local/bin/wine that executes box64 $HOME/wine/bin/wine "$@".
Unfortunately, when I run wine wineboot (with BOX64_LOG=1), I am still encountering errors related to ntdll.so failing to load due to missing _time64 symbols. The Box64/Box32 log shows the following (key parts):

[BOX64] Box64 arm64 v0.3.5 b1381125 with Dynarec built on May 20 2025 23:51:00
...
[BOX32] Using Box32 to load 32bits elf
...
[BOX32] Using emulated /home/hallotanakaz/wine/bin/../lib/wine/i386-unix/ntdll.so
[BOX32] Error: Symbol __prctl_time64 not found, cannot apply R_386_JMP_SLOT 0x40091000 (0x6036) in /home/hallotanakaz/wine/bin/../lib/wine/i386-unix/ntdll.so
[BOX32] Error: Symbol __gmtime64 not found, cannot apply R_386_JMP_SLOT 0x40091004 (0x6046) in /home/hallotanakaz/wine/bin/../lib/wine/i386-unix/ntdll.so
[BOX32] Error: Symbol __stat64_time64 not found, cannot apply R_386_JMP_SLOT 0x40091024 (0x60c6) in /home/hallotanakaz/wine/bin/../lib/wine/i386-unix/ntdll.so
[BOX32] Error: Symbol __fstatat64_time64 not found, cannot apply R_386_JMP_SLOT 0x40091034 (0x6106) in /home/hallotanakaz/wine/bin/../lib/wine/i386-unix/ntdll.so
(Many similar _time64 symbol errors follow)
...
[BOX32] Error: relocating Plt symbols in elf ntdll.so
[BOX32] Error initializing needed lib /home/hallotanakaz/wine/bin/../lib/wine/i386-unix/ntdll.so
wine: could not load ntdll.so: Cannot dlopen("/home/hallotanakaz/wine/bin/../lib/wine/i386-unix/ntdll.so"/0xXXXXXXXX, 2)

Use code with caution.
It seems the Box64 version I built (v0.3.5 b1381125) might still have issues resolving these _time64 symbols required by Wine 9.0.0's i386 ntdll.so when running via Box32 on my ARM64 Crostini (Debian Bookworm) system.
Could you please advise if:
A newer specific version/commit of Box64 is known to address these _time64 symbol issues for Box32?
There are any specific cmake options or environment variables I should be using with this version to improve _time64 symbol compatibility?
This might be related to the glibc version on the ARM64 host (Debian Bookworm's libc6 is 2.36-9+deb12uX)?
I am willing to try building the latest master branch of Box64 if that's recommended.
Thank you for your time and any guidance you can provide.

All outputs: https://paste.ee/p/H0UOOlIW

[ptitSeb]

Ah, more *_time64 wrapped function missing. I'll add them, but can you create a ticket in box64 github for that?

[ikumayuki]

I made that @ptitSeb

[ptitSeb]

Yeah, I have seen. Thx. I'll have a look at this soon-ish