Skip to content

Debug - #1683

Closed
NicolasHug wants to merge 1 commit into
meta-pytorch:mainfrom
NicolasHug:aeljnfeaf
Closed

Debug#1683
NicolasHug wants to merge 1 commit into
meta-pytorch:mainfrom
NicolasHug:aeljnfeaf

Conversation

@NicolasHug

Copy link
Copy Markdown
Contributor

No description provided.

@pytorch-bot

pytorch-bot Bot commented Sep 2, 2026

Copy link
Copy Markdown

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/meta-pytorch/torchcodec/1683

Note: Links to docs will display an error until the docs builds have been completed.

❌ 35 New Failures, 2 Unrelated Failures, 10 Unclassified Failures

As of commit 63b354a with merge base 60db9be (image):

NEW FAILURES - The following jobs have failed:

UNCLASSIFIED FAILURES - DrCI could not classify the following jobs because the workflow did not run on the merge base. The failures may be pre-existing on trunk or introduced by this PR:

FLAKY - The following jobs failed but were likely due to flakiness present on trunk:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@meta-cla meta-cla Bot added the CLA Signed This label is managed by the Meta Open Source bot. label Sep 2, 2026
pytorchmergebot pushed a commit to pytorch/pytorch that referenced this pull request Sep 2, 2026
code is from `claude`, text is from me.

----

#193451 [broke TorchCodec](meta-pytorch/torchcodec#1683). The issue is exactly the BC-breaking note acknowledged in #193451:

> BC-breaking note: we make torch.ops.* call go through the same exception and warning translation as the rest of the library. This means that some exception translation that used to be handle by pybind is now handled natively. User-defined pybind exception translation is thus lost here

TorchCodec [throws `std::out_of_range`](https://github.com/meta-pytorch/torchcodec/blob/85a89e2bd9d806381c60564f136fea8047114287/src/torchcodec/_core/StableABICompat.h#L39-L47) and relies on pybind to convert that to an `IndexError`. With #193451, this `IndexError` is now a `RuntimeError`, which [breaks various tests](meta-pytorch/torchcodec#1683) in the TorchCodec CI, as they expect an `IndexError`. These are user-facing exceptions, so it's not something I can transparently handle as implementation detail.

This PR lets `std::out_of_range` propagate to an `IndexError`, preserving the previous behavior before #193451. This is probably not the holistic fix that `torch` really needs, but I'm blocked on TC until this is resolved. I can also submit a revert PR for #193451 if that's the preferred solution.

(I am considering two temporary workarounds in TC: meta-pytorch/torchcodec#1684 which pins the torch nightly, or meta-pytorch/torchcodec#1685 which skip the failing tests - neither are viable long-term workarounds!)

Pull Request resolved: #195713
Approved by: https://github.com/malfet, https://github.com/albanD
@NicolasHug NicolasHug closed this Sep 3, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Meta Open Source bot.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant