search

JuliaCI / rootfs-images

Rootfs images used in Base Julia CI
MIT License
4 stars 6 forks source link

Bump msys2 version to the latest release as an attempt to fix the weird ASLR issue #250

Closed gbaraldi closed 4 months ago

gbaraldi commented 4 months ago

This bumps the msys2 versions to see if this can fix the weird ASLR issue we've seen in windows builders. This does mean we will get very new compilers for windows, this is hopefully not an issue (glibc compatibility isn't a thing and we ship a very new libstdc++)

It's GCC 13.2 for reference so it should be the same version libstdc++ that we have in julia since at least 1.10

giordano commented 4 months ago

This bumps the msys2 versions to see if this can fix the weird ASLR issue we've seen in windows builders.

For what is worth, I find these comments fairly useless for whoever isn't you: readers have no idea of what are the "weird issues" you're observing, and if this does solve the issue it won't be useful for other people who are actually experiencing similar issues but will never be able to find this page since there's no reference to what the issue is.

staticfloat commented 4 months ago

We discussed it on the #ci-dev call; it is a reference to failures such as this one:

> C:\buildkite-agent\hooks\environment
-> source C:\buildkite-agent\hooks/environment.d/100_gr_env.sh
-> source C:\buildkite-agent\hooks/environment.d/101_julia_num_cores.sh
-> source C:\buildkite-agent\hooks/environment.d/500_windows_debugging.sh
# Received cancellation signal, interrupting
      0 [main] bash (1824) child_copy: cygheap read copy failed, 0x18034C408..0x18036E1A0, done 0, windows pid 1824, Win32 error 299
    413 [main] bash (1824) C:\Program Files\Git\bin\..\usr\bin\bash.exe: *** fatal error in forked process - ccalloc would have returned NULL
🚨 Error: Error setting up bootstrap: The global environment hook exited with status 3221225786
Keno commented 3 months ago

More breadcrumbs: https://github.com/msys2/msys2-runtime/commit/a14a0e542d3aa8e53d189c906205c92ad044f750

Keno commented 3 months ago

However, I don't think that's the same issue we're seeing because in our crash logs, the base is not 0x00

Keno commented 3 months ago

There's also https://github.com/msys2/msys2-runtime/commit/317fc51b3af5280d8d115cab17a502f7c37b204c which I don't believe we picked up with this bump