Conversation
💊 CI failures summary and remediationsAs of commit 937f4cc (more details on the Dr. CI page):
🕵️ 15 new failures recognized by patternsThe following CI failures do not appear to be due to upstream breakages:
|
| Job | Step | Action |
|---|---|---|
| Download PyTorch Test Reports | 🔁 rerun |
ci.pytorch.org: 1 failed
This comment was automatically generated by Dr. CI (expand for details).
Follow this link to opt-out of these comments for your Pull Requests.Please report bugs/suggestions to the (internal) Dr. CI Users group.
|
@ngimel has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
|
@gdankel, there are windows build errors now: and I'm seeing the same errors in my local linux build when linking binary tests |
|
Ok yes updating submodule can unearth new issues for builds not covered by Kineto CI. We need to add Windows build there. Let me take a closer look... |
|
What's the urgency? |
|
Fix for #59746. I guess if proper submodule update is hard, we could cherry-pick file descriptor fix to a branch and update to that. |
|
Usually linker errors on windows are caused by incorrect symbol visibility, inside pytorch we usually decorate with TORCH_API, but kineto doesn't have this macro. |
|
I think that's a good approach for small fixes if larger changes have landed since last update. |
|
@gdankel can a branch exist in kineto github repo, not connected to fbcode? If so, can you create a branch from 24d562065 (current kineto submodule in pytorch) + cherry-pick 897a49cfd fix? I don't think I have permissions in kineto repo. |
|
Ok will do. And let me give you permission as well. |
|
Updated another PR with new hash |
I think I have fixed this linker error in pytorch/kineto#282 |
Another try for #59755