ivberg

> > They seem to have the "__traceiter_sched_switch" function but there is ONLY this function > > Do you mean that the callstack has only one frame? Seems weird. >...

Yes, that is what we see with the last function being similar to what you mentioned (line 21)  What I am saying is that the CPU Sampling events are...

> Yep I think that-s the issue. Yep - that's it! Sorry if it was confusing. At least it's possible to gather sched_switch stacks at all, so we are halfway...

Also - I looked several times but could not find a way to easily distinguish CPU Sampling events in [perf_sample](https://perfetto.dev/docs/analysis/sql-tables#perf_sample) from the sched_switch events which also show in perf_sample. There...

I tried the following but it didn't seem to work as I only received CPU sampling events/stacks. I think the sched_switch config is good by itself as it worked separately...

Please hold - this may be outdated. I forgot the full history here including some manual steps, and I think this is from the v38 version. #635 was fixed AFTER...

Ok - indeed Perfetto v45 fixes the error. I am leaving the new scan howto and warnings here in case someone wants to fix (optional). Using https://github.com/google/perfetto/releases/download/v45.0/windows-amd64.zip unzipped to say...

I agree. I am confused by (1) as well since https://github.com/google/perfetto/commit/2a4f01d2fc828bfb7a80926586c22d8b004aa1e9 code looks like it does add "/ZH:SHA_256". My best guess is some sub-lib compiled into the .exe didn't have...

We are seeing about sharing the model directly. It seems the alignment issue could happen on multiple platforms. I happen to be testing / using ARM64 though.

Here is some docs from the linked issue being closed saying this is semi possible now: https://perfetto.dev/docs/case-studies/scheduling-blockages