Source: python-memray Version: 1.17.0+dfsg-1 Severity: serious On top of the reported autopkgtest failure #1126944, there is another one shadowing the former one. Some ecosystem change (pytest?) results in tests failing to import memray:
| 57s ________________ ERROR collecting tests/integration/test_api.py ________________ | 57s ImportError while importing test module '/tmp/autopkgtest.ONqy8a/autopkgtest_tmp/tests/integration/test_api.py'. | 57s Hint: make sure your test modules/packages have valid Python names. | 57s Traceback: | 57s /usr/lib/python3.14/importlib/__init__.py:88: in import_module | 57s return _bootstrap._gcd_import(name[level:], package, level) | 57s ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | 57s tests/integration/test_api.py:5: in <module> | 57s from memray import FileDestination | 57s /usr/lib/python3/dist-packages/memray/__init__.py:2: in <module> | 57s from ._memray import AllocationRecord | 57s E ImportError: cannot import name 'AllocationRecord' from 'memray._memray' (unknown location) It is not clear why this happens now and didn't happen earlier. I tried understanding it, but did not. What I did figure though was that if you append "tests/" to the pytest invocation, this one gets fixed and we can see #1126944 again. On that other one, I spent some effort recorded in the upstream bug tracker. At this time I suggest uploading libunwind with the incompatible C++ exception support disabled. Before we can do that, this failure needs to be fixed. Helmut
