RFR: 7903740: JMH: Perf event validation not working with skid options [v6]

[Aleksey Shipilev]

On Fri, 6 Sep 2024 11:33:36 GMT, Galder Zamarreño <galder at openjdk.org> wrote:

>> Fixes https://bugs.openjdk.org/browse/CODETOOLS-7903740
>
> Galder Zamarreño has updated the pull request with a new target base due to a merge or a rebase. The incremental webrev excludes the unrelated changes brought in by the merge/rebase. The pull request contains 15 additional commits since the last revision:
> 
>  - Add -q to avoid additional perf record messages
>  - Pipe through perf report --stats instead
>  - Merge branch 'master' into topic.validate-perf-event-without-modifier
>  - Check all events with perf report in a single command
>  - Remove previous approach
>  - Use trim to remove any empty spaces or carriage returns
>  - Do not make it quiet otherwise there's no output
>  - Try perf report from specific file
>  - Remove old approaches
>  - Explicitly define an output file for perf record validation
>  - ... and 5 more: https://git.openjdk.org/jmh/compare/5a95accf...069447e1

Well, I just went to my `m7g.16xlarge` instance, and:


% java -jar benchmarks.jar -prof perfasm -f 1 JMHSample_35_Profilers.Atomic
 
Profilers failed to initialize, exiting.
Perf event count not a number for one of events [cycles]: [Can't parse sample, err = -14
0x224c [0x30]: failed to process type: 9
Error:
failed to process sample
0
]


It works without the patch. I think checking code is being too smart for its own good? Reiterating my previous suggestion: let's not try to parse the recordings, and instead "just" parse the output `perf record -e incorrect-event-blah` for testing the events.

-------------

PR Comment: https://git.openjdk.org/jmh/pull/132#issuecomment-2371993194