In mergin-py-client we're getting segfaults in CI with the new geodiff release:
File "/opt/hostedtoolcache/Python/3.10.5/x64/lib/python3.10/site-packages/pygeodiff/geodifflib.py", line 542 in __init__
File "/opt/hostedtoolcache/Python/3.10.5/x64/lib/python3.10/site-packages/pygeodiff/geodifflib.py", line 501 in next_entry
File "/opt/hostedtoolcache/Python/3.10.5/x64/lib/python3.10/site-packages/pygeodiff/geodifflib.py", line 509 in __next__
File "/home/runner/work/mergin-py-client/mergin-py-client/mergin/report.py", line 162 in changeset_report
File "/home/runner/work/mergin-py-client/mergin-py-client/mergin/report.py", line 269 in create_report
File "/home/runner/work/mergin-py-client/mergin-py-client/mergin/test/test_client.py", line 1684 in test_report
GDB further points to GEODIFF_CE_newValue() function.
0x00007ffff51382a0 in GEODIFF_CE_newValue ()
from /opt/hostedtoolcache/Python/3.10.5/x64/lib/python3.10/site-packages/pygeodiff/libpygeodiff-2.0.0-python.so
#0 0x00007ffff51382a0 in GEODIFF_CE_newValue ()
from /opt/hostedtoolcache/Python/3.10.5/x64/lib/python3.10/site-packages/pygeodiff/libpygeodiff-2.0.0-python.so
#1 0x00007ffff5bffff5 in ?? () from /lib/x86_64-linux-gnu/libffi.so.7
#2 0x00007ffff5bff40a in ?? () from /lib/x86_64-linux-gnu/libffi.so.7
It looks like in geodifflib.py, few C functions didn't get arguments updated when contexts were added in #172 - with GEODIFF_CE_newValue being one of them. On my local machine there's no segfault, so I can't properly verify the fix (but there's nothing else looking suspicious there).
Also:
switched from nose2 to pytest for python tests so that we use the same tool everywhere and fixed few places where pytest complained about things
added tests for column_is_pkey in changeset reader test
In mergin-py-client we're getting segfaults in CI with the new geodiff release:
GDB further points to GEODIFF_CE_newValue() function.
It looks like in geodifflib.py, few C functions didn't get arguments updated when contexts were added in #172 - with
GEODIFF_CE_newValuebeing one of them. On my local machine there's no segfault, so I can't properly verify the fix (but there's nothing else looking suspicious there).Also:
column_is_pkeyin changeset reader test