enable more ASAN tests

[cyyever]

Recently, we are seeing some bugs found by ASAN such as #101400, I think enabling ASAN for more tests is necessary to catch more hidden bugs.

cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/101483

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

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

✅ No Failures

As of commit e380233:
💚 Looks good so far! There are no failures yet. 💚

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

[cyyever]

@pytorchbot label "topic: not user facing"

[malfet]

Looks like your changes to build.sh make build-asan.sh redundant. If that's the case, please delete it.

Also, please include in the PR description how much this increases runtime of the tests.
And how one makes sure that the same test would not run on multiple shards(which is always the danger with .cpp tests) By the way, @huydhn haven't you automated running cpp tests and sharing them, that renders all those conditions in test.sh useless?

If not, than perhaps most prudent thing to do would be to create separate shard for C++ tests

[huydhn]

Also, please include in the PR description how much this increases runtime of the tests. And how one makes sure that the same test would not run on multiple shards(which is always the danger with .cpp tests) By the way, @huydhn haven't you automated running cpp tests and sharing them, that renders all those conditions in test.sh useless?

If not, than perhaps most prudent thing to do would be to create separate shard for C++ tests

Let's gather the information about the increase in test time and create a separate shard if need. At the moment, running C++ tests via run_test.py runs them in parallel (but still locally within one runner as I haven't setup the part to gather test time to enable sharding across runners yet)

[cyyever]

A misaligned error in PackedEmbeddingBagWeight::prepack may be relevant.

[cyyever]

@huydhn The slowest ASAN test takes over 2 hours.

[huydhn]

@huydhn The slowest ASAN test takes over 2 hours.

The increase was about 20m in the slowest shard. So I think we could tolerate that. A follow-up task would be to re-balance these ASAN shards. There are 6 of them and the 1st and 2nd shards need to handle more load atm.

[cyyever]

@pytorchmergebot merge

[pytorchmergebot]

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here