Kirsten's to-do list

* Someday, download a full repo so that I can check that the patches I want to send can be applied to the HEAD. grrrrrrrr

stuff

  1. I can't build the HEAD because ghc-6.4.2 coredumps for some reason.
  2. I can't build my "ghc-hacking" tree because interface file version checking doesn't work for some reason.
  3. I can't download a clean copy of the dmd-anal branch because I get a darcs error message (bothered Simon about it)
  4. Thus I don't have a working build tree and so I can't test my ticky patch with Simon's suggested fix. Not sure how I got into this state.
  5. Life, void, pain, misery, etc.

the answer: I did a bad merge at some point on iface/BinIface.hs and that ruined everything. I can at least build ghc-hacking now. Thank goodness for small blessings.

Demand analysis

current status: going through each file and sanity-checking.

  1. Get code ready to check in:
    1. cleanup
      • files to go through:
        • basicTypes/Id.lhs
        • basicTypes/IdInfo.lhs
        • basicTypes/NewDemand.lhs
        • coreSyn/CoreLint.lhs
        • coreSyn/PprCore.lhs
        • iface/BinIface.lhs
        • iface/IfaceSyn.lhs
        • iface/LoadIface.lhs
        • simplCore/SimplEnv.lhs
        • simplCore/SimplUtils.lhs
        • simplCore/Simplify.lhs
        • stranal/DmdAnal.lhs
        • stranal/WorkWrap.lhs
        • stranal/WwLib.lhs
        • types/Coercion.lhs
        • types/Type.lhs
    2. better comments (i.e., examples for things that may not work right)
    3. testing
  1. Update Commentary
  1. Experiments:
    1. comparison with old strictness analyzer
    2. see whether new optimizations are helping
    3. remove special cases for build and see what happens
  1. Paper
  1. Modify worker/wrapper split
    1. so as to exploit the new analysis information
  1. Think about the right way to handle this coercion stuff
  1. Formal semantics?

Old

KirstenSandbox/GhcOldMac - building GHC on Mac OS 10.2.1 (the summary: don't do it.)