You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[MNT] cuda.core: refresh the frozen CUresult explanation table for CUDA 13.3 #2370
CUDA 13.3 added CUDA_ERROR_GRAPH_RECAPTURE_FAILURE (13.2 added no new driver error codes), so
the new code currently has no fallback explanation text. The error will start appearing once
graph re-capture lands (#2359).
Proposed approach
Regenerate the snapshot against CUDA 13.3 using the same process that produced the 13.1.1
table (helpers are exercised by tests/test_utils_enum_explanations_helpers.py / tests/test_cuda_utils.py).
If not already covered there: add a tripwire test comparing cuda.bindingsCUresult
members against the table keys, so future toolkit bumps flag missing entries automatically.
Consider adding the refresh to the release checklist for new CTK support.
References
Found during the CUDA 12.8 → 13.3 bindings vs. cuda.core gap sweep (2026-07-14)
changed the title [-][FEA] cuda.core: refresh the frozen CUresult explanation table for CUDA 13.3[/-][+][MNT] cuda.core: refresh the frozen CUresult explanation table for CUDA 13.3[/+]on Jul 14, 2026
Activity
[-][FEA] cuda.core: refresh the frozen CUresult explanation table for CUDA 13.3[/-][+][MNT] cuda.core: refresh the frozen CUresult explanation table for CUDA 13.3[/+]Symbioose commentedon Jul 16, 2026
Hi @leofang , I'd like to work on this one.
Will open a PR shortly!