Markus Stange

Thanks for the note! Yes, I never anticipated that the tests would be run on 32 bit platforms, so it's fun to see that Debian does this. Also, thanks for...

I'd prefer handling both the old and the new search path. Thanks! I actually thought I had implemented this long ago, but I guess I hadn't! I think ever since...

> In our use case we are profiling full Android OS builds, where binaries on device are stripped but unstripped versions can be sourced from build directory. So we would...

Do you have the /proc/pid/maps from when you saw the incorrect / corrupted state? I wonder if it's not corruption but something else going on.

Can you paste the program header table and the section table here? (I mean the output of readelf that displays those tables) Which program header is the section table in?...

I have not seen this error before, but it probably indicates that the profile JSON contains some bad data. Can you reproduce this? Can you attach the corresponding profile.json file...

Is it an empty file? The tar.gz file shows me an error when I unpack it.

Ah, yes, all the sample tables are empty. You can search for `samples":{"length":` in the JSON. So for some reason samply wasn't able to capture any samples. Is it possible...

Interesting! How long does the profiled process run for? Maybe it shuts down before the first sample is gathered?