Closed chmreid closed 4 years ago
@chmreid I see that you requested my review, but didn't assign the PR to me. I'm confused by this. Did you see this Slack thread? https://ucsc-cgl.slack.com/archives/C705Y6G9Z/p1585093339100200
@jessebrennan The problem is with the master
branch, nothing specific to this PR. I tried re-running Travis tests for master and they failed: https://travis-ci.org/github/HumanCellAtlas/dcp-cli/builds/654957449
We can pick up on this PR once master is passing.
@jessebrennan Tests are passing, can you give this a review now?
I would like the opportunity to review this. I'm currently not assigned so I'll hold off for now.
@hannes-ucsc I have assigned this issue to you, please review when you get the chance. Thanks!
@hannes-ucsc Thank you for the comments. I have addressed your feedback, and this PR is ready for another review. Would you please review it at your convenience? Thanks.
This addresses a permission denied error on NFS file systems that is caused by the
os.link
call. Addresses #519Previously we were catching a "too many links" exception, now we are also catching a "permission denied" exception and an "operation not permitted" exception. (Note that we may be able to remove the "operation not permitted" exception number from the list of exceptions that lead us to fall back to copying.)
Useful Links
Useful discussion of possible approaches to solving this better: https://bugs.python.org/issue37612
Useful info about NFS and Python's os module: https://web.archive.org/web/20100912144722/http://www.unixcoding.org/NFSCoding
Note: this PR does not (yet) have tests for the new error codes being caught/handled. Tests TBA.