Ned Batchelder

Thanks for getting this started! I think all the examples should be put in branch.rst, since they will be helpful for people looking at the HTML report also. You can...

I haven't looked into the details here, but an alternative to copying data files around is to use an absolute file path for `[run] data_file`.

With `parallel = True`, each process will append its pid and other information to keep the data files distinct. An absolute path will mean they all end up in the...

A thing that might be little-known about coverage configuration files: they will expand environment variables. That might help get an absolute path you can use.

Those numbers are concerning. Can you add `--debug=sys,core` to your coverage runs so we can double-check that the right core is being used?

Thanks. I'll have to try some local experiments with this repo to see what's going on. My own timing tests show a slight slowdown, but only ~5%.

Thanks for the clues. Also, for the getsentry reproduction. I can confirm this is down to coverage.py and the Python version somehow: ``` % tox -qe py3.11-common -- -k test_basic...

The bulk of the problem is due to commit 31f91f816244a89051b301f2e3536e5f7b951e3d. Resetting coverage to just before that commit, I get: **19.90s** instead of 40.41s. But this is still much slower than...

I think this has to do with a proliferation of code objects. The basis of sysmon is that it can disable firing an event after it's been fired once, reducing...

I measured the unique code objects incorrectly, but it's still a lot. Looks like many of the `/sentry_sdk/integrations/*.py` files have 40 distinct code objects.