Fix slicing bug in ensure_memoryview

[jakirkham]

Fixes #6448

The underlying issues seems to be related to slicing. This example demonstrates the issue below and has been included in the test suite.

from distributed.utils import ensure_memoryview

mv = memoryview(b"0123456789").cast("B", (5, 2))[1:-1]
mv2 = ensure_memoryview(mv)

assert mv == mv2

• Tests added / passed
• Passes pre-commit run --all-files

[toloudis]

should ultimately update the comment above this code too

[jakirkham]

@JacksonMaxfield, would you be able to try this out? 🙂

[evamaxfield]

In class, gimme an hour or so and will do!

[jakirkham]

No rush. Thanks! 🙏

[github-actions]

Unit Test Results

       15 files  ±  0         15 suites  ±0   6h 19m 34s ⏱️ - 1m 15s
  2 816 tests +  4    2 735 ✔️ +  3    79 💤 ±0  2 ❌ +1 
20 875 runs  +28  19 944 ✔️ +29  928 💤  - 3  3 ❌ +2 

For more details on these failures, see this check.

Results for commit e1ccade. ± Comparison against base commit 046ab17.

♻️ This comment has been updated with latest results.

[evamaxfield]

@jakirkham here are the checks for this branch install: https://github.com/AllenCellModeling/aicsimageio/runs/6600770595?check_suite_focus=true

The important tests are the test-upstreams and so far it looks like they are all passing 🟢

Cannot believe this is less than a single line change to make it work. Fun.

[evamaxfield]

That said, maybe wait for macOS builds to finish up but your branch is working so far.

[jakirkham]

Thanks for checking Jackson! 🙏 Glad to hear it works 😄

[jakirkham]

Trying to reproduce the issue myself so I can come up with a test. Tried to do

curl -LO https://open.quiltdata.com/b/aics-modeling-packages-test-resources/tree/aicsimageio/test_resources/resources/3d-cell-viewer.ome.tiff

But ran into some issues. Is there something else I need to do to download this file?

Edit: NVM think I figured this out

[evamaxfield]

Hah, yea sorry that is just the viewer and UI downloader.

[jakirkham]

Yeah sorry poked around a bit and found the right URL. Now have it reproducing for me locally. Going to dig into it a bit more

[jakirkham]

Ok found a test that covers a similar problem. Think it is likely the cause of the issue in the original repro, but it is a little difficult for me to parse fully atm. In any event, this test will fail without the change here, which should ensure we don't regress

[pentschev]

rerun tests

Note: rerunning gpuCI tests since those errors should be fixed by #6434

[jakirkham]

Restarted the failed GitHub Actions for good measure.

Edit: Though they looked like flaky tests.

[jrbourbeau]

Thanks for the quick fix @jakirkham and reporting the original issue @JacksonMaxfield. This LGTM, will merge once CI finishes

[jakirkham]

All of the tests passed except for the macOS test, which is still queued. Do we want to wait for that?

[jrbourbeau]

The previous failure for that build was a flaky test, so happy to merge in

[jakirkham]

Thanks James! Also thanks Jackson for testing 🙏