Nightly feedstock build failed

[github-actions]

Nightly feedstock build failure for TileDB-Py%20Feedstock%20Testing at https://dev.azure.com/TileDB-Inc/CI/_build?definitionId=5&_a=summary

[github-actions]

Nightly feedstock build failure for TileDB-Py%20Feedstock%20Testing at https://dev.azure.com/TileDB-Inc/CI/_build?definitionId=5&_a=summary

[jdblischak]

We thought this was test failure was fixed by TileDB-Inc/TileDB#5417 because the tiledb-py-feedstock builds succeeded for 3 consecutive nights after it was merged (#167 (comment)). However, the same test test_multi_index_two_way_query started failing sporadically again Friday and Saturday nights. There haven't been any new commits to TileDB since then.

Worse, this is causing the tiledb-py-feedstock 0.33.4 update PR to fail (conda-forge/tiledb-py-feedstock#256)

[jdblischak]

The build passed last/Sunday Sunday night. Nothing actually changed upstream in TileDB or TileDB-Py; the flaky test just happened to pass in all 3 linux-64 builds.

[github-actions]

Nightly feedstock build failure for TileDB-Py%20Feedstock%20Testing at https://dev.azure.com/TileDB-Inc/CI/_build?definitionId=5&_a=summary

[ihnorton]

I think we should try to enable libSegFault to get a stack dump here, because we don't have much to go on right now: https://stackoverflow.com/questions/18706496/can-one-use-libsegfault-so-to-get-backtraces-for-sigabrt

cc @kounelisagis

[github-actions]

Nightly feedstock build failure for TileDB-Py%20Feedstock%20Testing at https://dev.azure.com/TileDB-Inc/CI/_build?definitionId=5&_a=summary

[jdblischak]

This is strange. The nightly tiledb-py-feedstock build didn't run last night. The failure reported above was from Monday (the Azure status check simply reports the result of the most recent build attempt).

Last night's setup job was successful. And it pushed to the nightly-build branch, but no Azure build was triggered by TileDB-Inc/tiledb-py-feedstock@6d47bc7

[jdblischak]

In contrast, the nightly tiledb-feedstock build ran last night after being triggered by TileDB-Inc/tiledb-feedstock@12ee1da.

For now I manually triggered a tiledb-py-feedstock build against the nightly branch. Hopefully it was simply a fluke. We'll see what happens tonight.

[jdblischak]

Confirming this was fixed. Last night's tiledb-py-feedstock build used commit TileDB-Inc/TileDB-Py@1b90a9f, which was the fix from TileDB-Inc/TileDB-Py#2166. All the linux-64 builds passed because the test test_multi_index_two_way_query was skipped.