Ergonomic, hermetic, user-friendly software build system for Python, Go, Java, Kotlin, Protobuf, Scala, Shell. Pants lets you fearlessly scale up your codebase!

Pants

to confirm the product issue with global, can `--skip` the dep check

<@U0AALHEDR>: is it because you're using `extend()` ?

<@U06A03HV1>:  I'm grabbing a snapshot of the heap when it's in this thrashing state, going to take a look at it when it finishes loading.  Anything in particular I should look for?

Based on the way the heap looks, I'd say this is a memory leak

<@U0AALHEDR>: if you're using extend, you definitely need maybe_list there, or at least a constraint on the type

well, regardless of the method. the type on the right hand side needs to be consistent. and that's what maybe_list is for.

<@U083H6LRL>: winding back up to your "where is the right place to put that requirement so pants itself picks it up? is that something that happens at bootstrap time?" If you list the plugins in pants.ini, pants bootstraps them and adds them to its pythonpath as needed, ie when you change the list:  ```
[DEFAULT]

plugins: [
    'pantsbuild.pants.contrib.go==0.0.45'
  ]
```

It honors your `[python-setup]` and `[python-repos]`, so if locked down, you need to plop sdists in your custom repos/indexes.

<@U06A03HV1>:  maybe_list works for one section of code, but once  bootstrap_option_values.config_override is called to append again somewhere else, then it will throw error, because itself is a string though it should be a list

<@U06A03HV1>: and it is a list if run locally

point is to make sure how cli args are parsed and stored consistently locally and on CI. also tried to copy to nest1 but it did not take it very well. shall i try spin a ubuntu vm locally and try again?

<@U06A03HV1>: Looks like we're leaking analysis objects

Screen Shot 2015-09-08 at 5.08.36 PM.png

Looks like there's a giant HashMap that the NGSession holds on to

It's under our namespace but I don't see it on our github

<@U053HNE90>: hm... those should be in the weak reference `Cache`, unless they are strongly held by the current compile

<@U053HNE90>: in the pantsbuild/pants repo

can you confirm that those are held by the current compile, rather than by the cache?

<@U0AALHEDR>: i don't think the issue is linux/osx related.

the point is just that different codepaths to add to that option use different types. need more type enforcement.

<@U0AALHEDR>: is this patch posted somewhere?

Screen Shot 2015-09-08 at 5.17.47 PM.png

Seems like the NGSession itself has a handle on 25%

<@U06A03HV1>: restarted on CI before after clearing the cache, issue persists. patched is at <https://rbcommons.com/s/twitter/r/2773/diff/1#index_header>