Julian Gilbey

Hi @bluetech, thanks for your detailed comments! > > python3.12 -m pytest /tmp/pytest-order-subtests > > I am not sure why the rootdir ends up as `/tmp`. From looking at the...

Actually, I'm not sure if that patch is the cause now; `get_common_ancestor` is now called with the invocation directory as a fallback in place of `Path.cwd()`. So I'm a bit...

Ah, figured out the comment about `--rootdir`. The following succeeds: ``` $ cd /tmp $ python3.12 -m pytest --rootdir=/tmp/pytest-order-subtests /tmp/pytest-order-subtests ```

Here's a suggestion: change https://github.com/pytest-dev/pytest/blob/12e061e2e838fb2c17d54df40a4a11aeb8723c6b/src/_pytest/config/findpaths.py#L210-L215 to simply read: ```python if rootdir is None: rootdir = ancestor ``` The reasoning is this: if pytest is called as `python -m pytest dir...

#263 as well :)

Thanks! It seems to be slightly inconsistent which test it crashes at (and occasionally it succeeds on the whole set of tests). (And it consistently succeeds with Python 3.11.) I've...

Adding a bit more to my previous comment. I've just tried running `python3.12 with debug options: ``` (sid_armhf-dchroot)jdg@abel:~/python-bytecode-0.15.1$ python3.12 -d -X dev -m pytest tests ============================= test session starts ==============================...

I tried running with the failing test excluded: ``` (sid_armhf-dchroot)jdg@abel:~/python-bytecode-0.15.1$ python3.12 -m pytest tests -k "not test_huge_code_with_numerous_blocks" ============================= test session starts ============================== platform linux -- Python 3.12.1, pytest-7.4.3, pluggy-1.3.0 rootdir:...

> > Backtrace stopped: previous frame identical to this frame (corrupt stack?) > > This seems to be another cue for frame stack corruption. I've had these in the past...

Hi @MatthieuDartiailh, thanks for taking a look at this! I'm not sure I fully understand your request, I'm afraid, but I'll give it my best shot. Here is what I've...