See PR46515 for the rational but generally, we want to *really* abort
not gracefully shut down.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Unit Tests
Event Timeline
abort() does not flush stdout, but here we are printing to stderr which is not buffered so there is no need to flush.
stderr is unbuffered by default. Are we sure this behavior is not changed when the output is redirected into a file, as for example during execution in a batch job? What were the worries about having the fflush there?
I went with this: https://reviews.llvm.org/D83963#2156612
If you think flush is needed, feel free to add it, it doesn't cost us anything.
This change broke a number of libomptarget tests. Adding --crash to the %libomptarget-run-fail-* definitions in openmp/libomptarget/test/lit.cfg fixes that for me. Any objections?
not expects a non-zero exit not a signal. Adding --crash adjusts the expectation.
I would've just pushed the fix, but I was concerned that maybe others weren't seeing this somehow.
clang-tidy: error: unknown type name 'DeviceTy' [clang-diagnostic-error]
not useful