[kristjanvalur]

Issue #93592 describes how a coro.throw() will wake up the coroutine, but the frame chain is broken, making it
impossible to get a proper traceback. This impedes debuggers when they break on thrown exceptions, among other things.

The problem is observed in versions 3.7-3.10 but is fixed in 3.11, since there appears to have been structural changes made in frame objects in that version, and the necessary re-write of the generator code seems to have (accidentally?) fixed the issue.

A regression tests is provided, which could also be applied to the 3.11 and main branches.

• Issue: Weirdness with traceback when resuming coroutines with coro.throw() #93592

[kumaraditya303]

Please the PR against main branch and then it will be backported to older branches. Thanks!

[kristjanvalur]

@kumaraditya303 PR doesn't apply to main. See above, issue appears to have been accidentally fixed in 3.11, without having been backported.

[kumaraditya303]

PR doesn't apply to main. See above, issue appears to have been accidentally fixed in 3.11, without having been backported.

Right, re-opened Sorry for closing your PR, apparently the PR title did not contain [3.10] hence I closed it as we backport changes from main to older branches.

[kumaraditya303]

@kristjanvalur You can add a news entry with https://blurb-it.herokuapp.com/

[kristjanvalur]

This has been sitting here for a bit. I wonder if we could request a review from someone else?

[markshannon]

Thanks @kristjanvalur.

Would you mind porting the regression test to main, to make sure that this stays fixed?

[kristjanvalur]

Not at all. I'll create a separate PR.