Repository navigation
Fix issue 8370 - #8376
Fix issue 8370#8376
Conversation
|
Many thanks for the PR!
I'm not sure |
|
gpuci run tests |
|
(Marking as RFR because the test fail is a CI issue, not an issue with this PR) |
… whether a copy was necessary or not.
…er a copy is needed
|
Resolved issue picked up by an automated test: if the source array is mutable, then a simple ravel() should not make it read-only. |
|
Ready for review. |
|
Note: the algo modification is not a big deal:
That's all. |
|
/azp run |
|
gpuci run tests |
|
Azure Pipelines successfully started running 1 pipeline(s). |
gmarkall
left a comment
There was a problem hiding this comment.
Many thanks for the updates and your continued efforts! There are a couple of comments on the PR.
In addition, in test_readonly_after_ravel and test_readonly_after_flatten, there is no assertion that checks that an expected output is produced - could you also assert that cfunc and pyfunc both produce the same return value please?
|
Also, apologies for the delay in this round of review. |
…ause different version of Numba return different way to say that you are trying to modify a read-only array. * added an assertion to check the Numpy and Numba return the same results.
gmarkall
left a comment
There was a problem hiding this comment.
Many thanks for the quick update - just a very small comment on the diff - would you be happy to make this change, then this should be good to approve?
|
Many thanks - just waiting on CI before I approve. |
|
gpuci run tests |
gmarkall
left a comment
There was a problem hiding this comment.
Many thanks for the PR and updates! I think this looks good now.
Fixes #8370
New unit tests were created, and the solution added, as suggested by @gmarkall
All unit tests pass locally that would exercise ravel() and flatten()
CUDA tests may need to be run, but I simply cannot download the CUDA drivers from Nvidia today for some reason, so maybe a CI test would be more appropriate.