Spurious LINE event when tracing `asyncio.tasks.sleep`

[markshannon]

See nedbat/coveragepy#1368 (comment) for report.

Although this occurs in asyncio.tasks.sleep I suspect that this this is more generally related to entering coroutines and generators.

[markshannon]

Reproducer:
nedbat/coveragepy#1368 (comment)

[nedbat]

Is there any chance of this being addressed in time for 3.11b1?

[markshannon]

Sorry, I was unable to reproduce this, so I couldn't make for a test for it nor fix it.

I think I know what the issue is now.
There is a spurious LINE event before the call, but pure Python tracing ignores LINE events that occur before the CALL event because f_trace is unset until set by the global trace function.
C tracing functions see this LINE event.

[nedbat]

Where does this leave us? Is this something that will change in CPython, or does coverage.py need to adapt?

[markshannon]

I'm fixing it. It just didn't make it into 3.11b1, obviously. It will be in 3.11b2

[markshannon]

#92722

[markshannon]

@nedbat thanks once again for finding yet another tracing bug.