| FazBrowse GitHub Viewer | Trending | | Home |
| Tools: [Download Repo ZIP] [Original HTTPS Page] |
Sorry, something went wrong.
The CI runners now have LongPathsEnabled=1 (nv-gha-runners/vm-images#241), so remove the skip of pip install and all_must_work tests for py3.14t on Windows. Add a fail-early step to catch runners that lack the setting. Closes NVIDIA#1820 Made-with: Cursor
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
Sorry, something went wrong.
|
/ok to test |
Sorry, something went wrong.
Disable Linux builds/tests and docs, filter Windows test matrix to py3.14t to verify LongPathsEnabled fix in isolation. Revert before merge. Made-with: Cursor
|
/ok to test |
Sorry, something went wrong.
This reverts commit a92790a.
|
/ok to test |
Sorry, something went wrong.
|
Sorry, something went wrong.
| echo "LongPathsEnabled = $val" | ||
| if ($val -ne 1) { | ||
| echo "::error::LongPathsEnabled is not set to 1 (see issue #1820)" |
There was a problem hiding this comment.
| echo "LongPathsEnabled = $val" | |
| if ($val -ne 1) { | |
| echo "::error::LongPathsEnabled is not set to 1 (see issue #1820)" | |
| if ($val -ne 1) { | |
| echo "::error::LongPathsEnabled is $val but should be set to 1 (see issue #1820)" |
No reason to print the value if it's 1, nor is there a reason to be so cryptic in the error message: just combine saying what it is with what is expected.
Sorry, something went wrong.
There was a problem hiding this comment.
This is for a CI log, having a confirmation about LongPathsEnabled is useful (especially for LLMs to build context). Why do you not want to show that one line?
The output with the code as-is:
LongPathsEnabled = 0 ::error::LongPathsEnabled is not set to 1 (see issue #1820)
Seems very clear to me:
The solution needs some explanation (unless you know about it already), so the pointer to the issue is a very efficient way to helpful.
Sorry, something went wrong.
There was a problem hiding this comment.
I don't follow the LLM argument here. Is an LLM concerned about non failure case? Why does it need to have context in the boring, successful case. The context is there from the error message.
I personally despise output fluff like "build done" or "things went well", or just echoing information just because maybe it'll be useful someday.
Unless it's notable, actionable information, like a failure or actionable warning, then there should be no output. I'm not saying there are no exceptions to that, but this to me is a clear case of don't add any more spam to CI logs than necessary.
Sorry, something went wrong.
| Back | FazBrowse Home | New Git URL |
Closes #1820
The CI runner images now have LongPathsEnabled=1 (nv-gha-runners/vm-images#241), so the nvidia-cutlass wheel installs successfully under free-threaded Python on Windows even though the longest path (266 characters) exceeds the default MAX_PATH limit.
Changes