Status fields:
| creation_ts: | 2010-11-11 19:53 |
|---|---|
| component: | jit |
| version: | default branch |
| rep_platform: | powerpc64 |
| op_sys: | Linux |
| bug_status: | RESOLVED |
| resolution: | WORKSFORME |
| reporter: | stefan@complang.tuwien.ac.at |
Since I've had to move my testing to a different powerpc64 machine, I'm getting SIGILL on code that looks completely valid, so I guess that something must be wrong with instruction cache flushing. Unfortunately, gdb on said machine is flaky at best, so I can only look at core dumps. It happens with "cacao hello", but not with "cacao" only.
Doesn't happen on our PPC64 machine. Maybe differing cache line sizes?