[PATCH] D34444: Teach codegen to work in incremental processing mode.

[Richard Smith - zygoloid via Phabricator via cfe-commits]

rsmith added a comment.

This makes sense to me, but I'd like John's input on whether this is a reasonable facility for CodeGen to have, and whether this is sufficient for (for example) inline function definitions to be emitted at the right times into the right `Module`s. (The tests for this appear to be in https://reviews.llvm.org/D34059.)


Repository:
  rL LLVM

https://reviews.llvm.org/D34444