I'm excited to pick up some graph garbage collecti...
# development
f
I'm excited to pick up some graph garbage collection work again this weekend. Running
pants dependents
for me now takes 1 min and consumes 3.8 GiB in pantsd (starting from no pantsd, but warm local cache) 😳 I have a decent understanding of why the dependencies requests take so long now, but do we have any quick insights into what might consume so much memory here? I currently have 8070 targets, which doesn't seem like that much. I'll be interested to see how much GC can actually prune here
https://github.com/pantsbuild/pants/pull/19513 picks this GC stuff up where I left off, although I'm kinda disheartened. Clearing nodes from the graph while Pants is actually executing rules seems to just make it run the rules again😄 I guess this is why reviewers in original PR had suggested such long times for thresholds (like 10 minutes). Trouble is, this isn't really helpful for my actual use case of
pants dependencies ::
exceeding 4 GB in a single run in under one minute. "Unreachable" nodes are supposed to be irrelevant to currently executing rules, right?