Skip to content

compile back to hard scope when closure compile/build requested#4392

Merged
enebo merged 2 commits intomasterfrom
compile_more
Jan 11, 2017
Merged

compile back to hard scope when closure compile/build requested#4392
enebo merged 2 commits intomasterfrom
compile_more

Conversation

@enebo
Copy link
Member

@enebo enebo commented Dec 16, 2016

tbh I am super bothered I cannot figure out a test case to at least occasionally fail. I explain the basic issue in #4304 (more specifically #4304 (comment)). Summarized as two threads both decide to build something and one is a child closure and the other is a parent of that closure. We get a sequence like:

  1. child starts
  2. parent starts
  3. parent accesses childs CFG
  4. child changes CFG
  5. parent accesses childs CFG
  6. BB is missing

This code will ask hard scope to full build itself for blocks and the two entry points to building on IRScope will built the closures before the parent. It does not appear to infinite loop so yay? I could not see anything bust with this change and I think the intent is clear but I am unsure in best particular impl to use. Part of it is based on children wanting parent info for passes vs parents wanting children info for passes.

@subbuss I think you have best knowledge of these expectations and @headius I would like your thoughts as well.

@enebo
Copy link
Member Author

enebo commented Jan 11, 2017

Giving it a shot...

@enebo
Copy link
Member Author

enebo commented Jan 17, 2017

#4424 also seems to be this issue. If so we should resolve #4424 as a duplicate of #4304

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant