[PATCH] Preliminary support for dynamically loadable coff objects

[Andy Ayers]

Ok, thanks. Will refactor to split out the architecture-specific bits.

On the testing front -- I started trying to load up one of our generated .objs via rtdyld and ran into some assumptions in the base loader code. In particular the object we generate has a zero-sized .data, and the loader code is a bit too eager to get at the contents of .bss. But our objs are kind of degenerate; I don't know if the intent is to have this code work with arbitrary objects or just the plausible ones you'd see from a compiler or assembler where zero-sized sections probably get squashed.


REPOSITORY
  rL LLVM

http://reviews.llvm.org/D7793

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/