More generally, the issue happens with all exceptions raised by other threads.

Run the code below and interrupt with ctrl+c.

# If part of a file, potentially more code before the loop. Traceback says -1 regardless.
for i in range(30000000):
  if not i%1000000:
      pass  # Or any other operation here that is faster than iterating a million i values.

With the following traceback:

Traceback (most recent call last):
 File "/home/panta/issue.py", line -1, in <module>
KeyboardInterrupt

The issue disappears as soon as we add just a bit more to the loop

for i in range(30000000):
  if not i%1000000:  # The 3.10 traceback points to this line.
      pass
  pass  # The 3.11 and 3.12 tracebacks point to this line.

Traceback (most recent call last):
File "/home/panta/issue.py", line 4, in <module>
  pass
KeyboardInterrupt

The issue persists whether running from the shell or as a file or even using exec. It works the same inside functions/methods. It exists in 3.11 and 3.12 but not 3.10, across Windows and Linux.

• CPython versions tested on: 3.10.6 (no issue), 3.11.1 (issue), 3.11.4 (issue), 3.12.0rc1 (issue)
• Operating system and architecture: Windows 10 and Ubuntu 22.04.2 LTS (WSL)

Linked PRs

• gh-107901: Fix missing line number on BACKWARD_JUMP at the end of a for loop #108242
• [3.12] gh-107901: Fix missing line number on BACKWARD_JUMP at the end of a for loop (GH-108242) #108275
• [3.11] gh-107901: Fix missing line number on BACKWARD_JUMP at the end of a for loop #108326
• gh-107901: add the HAS_EVAL_BREAK instruction flag #108375
• gh-107901: synthetic jumps which are not at end of loop no longer check the eval breaker #113721
• gh-107901: jump leaving an exception handler doesn't need an eval break check #113943
• gh-107901: duplicate blocks with no lineno that have an eval break and multiple predecessors #113950
• gh-107901: guarantee_lineno_for_exits is not doing anything #114008
• gh-107901: compiler replaces POP_BLOCK instruction by NOPs before optimisations #114530
• gh-107901: make compiler inline basic blocks with no line number and no fallthrough #114750