search

open-rmf / rmf_ci_templates

Templates for workflows and CI
1 stars 4 forks source link

Address sanitizer fails on rolling because of uncrustify #12

Open luca-della-vedova opened 3 months ago

luca-della-vedova commented 3 months ago

Before proceeding, is there an existing issue or discussion for this?

Description

As documented in https://github.com/open-rmf/rmf_ci_templates/pull/10#issuecomment-2130315925, it seems that the asan action fails on rolling because of uncrustify, possibly due to the fact that we use a custom style file throughout Open-RMF, the fact that the uncrustify version changed in 24.04 or both. What is even more puzzling though, is that the normal build and test on Rolling / 24.04 succeeds, so it is not purely an uncrustify issue but seems to be related at the asan action.

mxgrey commented 3 months ago

This is a really bizarre problem. In all honesty there's no reason for us to be running uncrustify during the asan test anyway.

While it's frustrating to simply cover up a problem without understanding it, I don't see much value in spending time to figure out what inexplicable combination of environment settings is causing this illogical outcome. We're actively working to retire this codebase, and I'd rather we focus on that rather than getting bogged down in parts of its CI that shouldn't even matter.

I would suggest we try to suppress uncrustify tests during asan runs using cmake test labels. I think if we give the uncrustify tests a label and just exclude that label during asan runs, we can just sidestep all this weirdness.