search

igraph / rigraph

igraph R package
https://r.igraph.org
532 stars 200 forks source link

Flaky sanitizer builds #1311

Open krlmlr opened 3 months ago

krlmlr commented 3 months ago

What happens, and what did you expect instead?

Only about 3 out of 10 runs for the sanitizer actually succeed, most others die with a segfault. I believe this has started when we turned on running the sanitizer for examples.

To reproduce

https://github.com/igraph/rigraph/actions/workflows/build-and-check.yml

System information

No response

szhorvat commented 3 months ago

Could be due to randomness in some examples? Segfault still indicates a bug though, this should be checked ...

krlmlr commented 3 months ago

The segfaults sometimes occur during an odd stage, though -- during package installation.

Antonov548 commented 3 months ago

I have investigated it a bit. The problem is occurred after running this pipeline with introduced memory leak. https://github.com/igraph/rigraph/pull/1286

Then it creates cache of the GHA and when it's used segfaults happens during installation. So, cleaning of the cache for GHA fix this problem. I guess we should just dissable caching for this step. Not sure how caching is working with the steps which use docker containers.

krlmlr commented 3 months ago

I don't see a cache created on GHA.

Antonov548 commented 3 months ago

I don't see a cache created on GHA.

I just removed it to test more this theory. But locally I don't see segfault in the docker container.