FIX: Skip refcheck in ArraySequence construction/extension#719
FIX: Skip refcheck in ArraySequence construction/extension#719effigies merged 1 commit intonipy:masterfrom
Conversation
Codecov Report
@@ Coverage Diff @@
## master #719 +/- ##
======================================
Coverage 89.1% 89.1%
======================================
Files 93 93
Lines 11468 11468
Branches 1991 1991
======================================
Hits 10218 10218
Misses 911 911
Partials 339 339
Continue to review full report at Codecov.
|
effigies
left a comment
There was a problem hiding this comment.
Okay, I've looked through this file, and tried to think through the ways that you might hit this function, and it seems like the ndarray object should be pretty isolated, so the refcount doesn't seem to be going up for a reason we need to guard against.
So I'm inclined to merge.
@matthew-brett @MarcCote This is in code you two co-wrote, so if you see a reason we need to seriously worry about breaking other objects sharing a buffer, then please make some noise.
I'll plan to merge on Friday.
|
I defer to @MarcCote on this one ... |
|
After looking at the resize documentation, I'd say let's merge this. It should be fine. |
|
Thanks for the review. |
I stumbled into this in the course of my work on pyAFQ.
For testing, part of the code reads data from a trk file. This seems to cause some issues in this method, but only on 3.7: https://travis-ci.org/arokem/pyAFQ/jobs/483151878. Unfortunately, I was not able to reproduce these errors locally on my machine...
After reading about what this might be, and implementing the proposed change, these tests are now passing on the Travis build for that branch, though.
Any objections to this?