Hello. I've observed some unexpected behaviour around task cancellation/retries and I'm trying to work out if I've misconfigured pants or not. In particular: I'm noticing that any file system change will cause any running tasks to be retried, even if the file system change doesn't affect the running tasks. For instance, the task might be a long-running test or a server (i.e. runs forever), and the change might be: • touching any file without changing contents (equivalent of

touch path/to/file

) • changing a file that the running task(s) do not (transitively) depend on For either of these, I'd expect the running tasks to keep running, and it's annoying that fiddling on some unrelated part of a codebase forces things to restart. Is this the expected behaviour?

i have exactly the same issue. tried different pants versions with no change in behaviour

this is tracked here: https://github.com/pantsbuild/pants/issues/10705 this is not the intended behavior, it’s just that the current implementation of dirtying is aggressive… there is the beginning of a design sketched out on the issue

Oh, cool. Thanks (embarrassingly, I see I'd even commented on that issue before, and forgotten about it 😅 )