rayon-rs/rayon

par_bridge_recursion test sometimes hanging indefinitely

decathorpe opened this issue · 3 comments

decathorpe commented

I have noticed that there is one test that sometimes just gets "stuck" and hangs indefinitely (an hour or longer, depending on how patient I am with cancelling the build):

     Running `/builddir/build/BUILD/rayon-1.9.0/target/rpm/deps/par_bridge_recursion-fe4a8a5c3a55045a`
running 1 test
test par_bridge_recursion has been running for over 60 seconds

For example, the total build time for rayon 1.9.0 was 4-5 minutes on x86_64, i686, s390x, but on ppc64le, this test has been stuck for almost two hours now.

Build environment:

  • Fedora Linux / Rawhide on ppc64le
  • Rust 1.76.0
  • distro-wide default RUSTFLAGS: -Copt-level=3 -Cdebuginfo=2 -Ccodegen-units=1 -Cstrip=none
cuviper commented

Maybe #1119 needs a closer look?

decathorpe commented

PS: I just remembered that this might actually be not specific to ppc64le. I have seen this test hang on other architectures too, but it might just be most likely to happen on ppc64le because that is our slowest architecture and builders are often under high load.

cuviper commented

I know it's hard in a koji builder, but the most useful thing would be a stack trace of all threads of the hung process.