"No space left on device" regression with 2.18.1

[cary-ilm]

Description of the issue

The CodeQL c-cpp workflow on the OpenEXR repo began failing on July 27. We use the default configuration, and runs have traditionally taken 4-12 minutes, but without significant changes to our code, runs began taking 1 hour and failing with the disk filling up. I reverted the repo back to the commit from the last successful run and it failed, indicating it's not likely something on our end.

I managed to revert the version of CodeQL in the workflow to 2.18.0, and everything is back to working smoothly.

Any clues what might be causing this?

The log from the last successful run is here. Log from the first failed run is here.

The PR that address the problem for OpenEXR by reverting to 2.18.0 is here.

[geoffw0]

Hi @cary-ilm , I have confirmed there is a problem with the performance (both time and space) of alias analysis computations on this database. We'll let you know here when there's a solution.

[jketema]

Hi @cary-ilm,

We merged a fix. The fix should be part of CodeQL 2.19.3, which will be released in a few weeks.