f_trace gets call on different lines each loop

[MegaIng]

BPO	42286
Nosy	@MegaIng
Files	f_trace_wrong_line.py

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = None
created_at = <Date 2020-11-07.22:19:52.971>
labels = ['interpreter-core', 'type-bug', '3.9']
title = 'f_trace gets call on different lines each loop'
updated_at = <Date 2020-11-07.22:19:52.971>
user = 'https://github.com/MegaIng'

bugs.python.org fields:

activity = <Date 2020-11-07.22:19:52.971>
actor = 'MegaIng'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Interpreter Core']
creation = <Date 2020-11-07.22:19:52.971>
creator = 'MegaIng'
dependencies = []
files = ['49580']
hgrepos = []
issue_num = 42286
keywords = []
message_count = 1.0
messages = ['380527']
nosy_count = 1.0
nosy_names = ['MegaIng']
pr_nums = []
priority = 'normal'
resolution = None
stage = None
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue42286'
versions = ['Python 3.9']

[MegaIng]

When using a local trace function, sometimes the line event is issued at the wrong time. As can be seen from the example file f_trace_wrong_line.py, each loop iteration we get either 24 or 23, apparently in a fixed rhythm.

While this might seem harmless, it appears that the line event for 23 (In this example) is issued at a time where the state of the stack is not valid to switch line. I noticed this because I wanted to implement a context manager that skips to a specific line, but this broken with mysterious error message TypeError: 'NoneType' object is not callable which I think is caused by the with-line not being fully executed and the stack still having the return value of __enter__ on it.

Here the intro line of the interpreter:

Python 3.9.0 (tags/v3.9.0:9cf6752, Oct 5 2020, 15:34:40) [MSC v.1927 64 bit (AMD64)] on win32

[iritkatriel]

I always get line 25 on main (3.14) on a Mac.

[iritkatriel]

CC @gaogaotiantian

[gaogaotiantian]

In recent versions we have made plenty of improvement (changes) related to line number calculation. The tracing mechanism has been completely different as well. I can't repro this on 3.14 either. Actually I can only repro this on 3.9 (something in 3.10 fixed it). As 3.9 does not take bug fixes anymore, there's really nothing we can do here. So I'll just close this.