Open LoopedBard3 opened 8 months ago
FYI: @matouskozak
I see, the change in #2367 returned error if it wasn't successful, while my refactor in #2393 just uses the ToBuildResult
method, which handles it differently for warnings (I thought it was the same). https://github.com/dotnet/BenchmarkDotNet/blob/8631651a33f46c68e07972bbdf72eae7736fceb0/src/BenchmarkDotNet/Toolchains/DotNetCli/DotNetCliCommandResult.cs#L37-L40
That file is 5 years old at this point, is the exit code 1 from only a warning still an issue on newer sdks? Maybe we can remove that check. #2367 effectively did that anyway by skipping that check.
We recently found a MonoAOTLLVM performance test run that failed to build but still ran and produced results when we expected it to fail instead. Here is part of the log that demonstrates this:
This occurred with dotnet/performance commit hash 044ae265469659d6ef7bf65c8af040652fe7ef31, dotnet/runtime commit hash 9dc6ea62a4d195ae4559f4609a56933c61889756, and BenchmarkDotNet version 0.13.13-nightly.20240213.132.
I was able to repro the fail but still run by checking out the specified dotnet/performance repo hash, cd into the scripts folder, and running:
sudo python3 benchmarks_local.py --commits 9dc6ea62a4d195ae4559f4609a56933c61889756 --run-types MonoAOTLLVM --filter *Span.IndexerBench.CoveredIndex2* *WriteReadAsync*
. I did this in WSL on an Ubuntu 22.04.3 LTS instance.Potentially related to https://github.com/dotnet/BenchmarkDotNet/issues/2311 and found during investigation of https://github.com/dotnet/runtime/issues/98967.