Adam Williamson

ok then, let's try them both as arrays...

ugh. no. I'm pretty sure it was right the first time: first def is an array, second is a string. back to that.

Unfortunately the build environment I'm using doesn't preserve the coredump, I'll see if I can reproduce it differently and get the coredump.

OK, here's the traceback. I guess it points to rasterio or gdal being the issue? It's hitting one of the asserts added to `gdalsievefilter.cpp` in https://github.com/OSGeo/gdal/commit/7de0934c70c0b4d57d2b86c40b02b611efa74bac . ``` #0 __pthread_kill_implementation...

confirmed the test does not crash when run against gdal 3.9.1, thanks a lot.

Without this, if you try and run the test suite on Python 3.13, you get a traceback like this: ``` _______________ ERROR collecting skimage/draw/tests/test_draw.py _______________ /usr/lib/python3.13/site-packages/_pytest/runner.py:341: in from_call result: Optional[TResult]...

looks like there are three test failures after this, at least one of which is also due to the indentation change. working on those.

hmm, this may not be capturing all lines of the description, by the looks of it...looking into why...

agh, yeesh, now I understand exactly how the regex works, it's actually very difficult to fix perfectly. it relies on the indentation, without it you can't really capture the last...

The CI failure is genuine: `C:\hostedtoolcache\windows\Python\3.11.9\x64\Lib\site-packages\skimage\measure\_regionprops.py:docstring of skimage.measure._regionprops.regionprops:210:Definition list ends without a blank line; unexpected unindent.` well, yes, indeed I did make it so the list of properties doesn't exactly...