1. 20 Nov, 2000 9 commits
  2. 19 Nov, 2000 2 commits
  3. 17 Nov, 2000 5 commits
    • simonmar's avatar
      [project @ 2000-11-17 17:03:17 by simonmar] · 71b3f476
      simonmar authored
      add a few missing cases
    • simonmar's avatar
      [project @ 2000-11-17 16:53:27 by simonmar] · 459e7bd4
      simonmar authored
      Results of today's hacking:
        - We can now execute expressions from the GHCi prompt.  However,
          a problem with the typechecker environment means that identifiers
          from outside the current module aren't resolved :-(
        - loading up a multi-module program in the interpreter seems to
          work.  Interpreting is kinda slow (ok, very slow), but I'm hoping
          it'll get better when I compile the interpreter w/ optimisation.
        - :set sort of works - you can do ":set -dshow-passes", for example
        - lots of bugfixes, etc.
    • sewardj's avatar
      [project @ 2000-11-17 13:33:17 by sewardj] · a5f9c20a
      sewardj authored
      Facilitate handling the situation where name of module /= name of file.
      This is handled in CompManager.summarise, which can inspect the source
      to see what the _real_ module name is, and act accordingly.
      All a bit messy -- needs a rethink.
    • sewardj's avatar
      [project @ 2000-11-17 11:59:59 by sewardj] · 22e83441
      sewardj authored
      Filter the HST and HIT passed to compile to only contain the downwards
      closure of the module being compiled.  This means that the instance-env
      construction machinery cannot pick up out-of-scope instances from HST.
      Don't know if it's necessary to filter HIT -- perhaps not.
      Instance env construction still doesn't work right, due to duplicates
      being picked up from interface files.
    • sewardj's avatar
      [project @ 2000-11-17 10:13:21 by sewardj] · f5f3b9d9
      sewardj authored
      Make file up-to-dateness checking work in batch mode.
  4. 16 Nov, 2000 11 commits
  5. 15 Nov, 2000 6 commits
    • simonpj's avatar
      [project @ 2000-11-15 17:07:34 by simonpj] · ea659be5
      simonpj authored
      I finally got tired of not having
      (Previously we called splitTyConApp_maybe,
       but it's a pain in the neck.)
    • simonpj's avatar
      [project @ 2000-11-15 17:07:17 by simonpj] · 894a5792
      simonpj authored
      The main thing in this commit is to make sure the desugarer
      looks a the right TyCon when desugaring a record update.
      Again, it's because in GHCI-land, it's possible that the
      some versions of the TyCon won't have the constructors;
      but the one in the FieldLabel will.
    • sewardj's avatar
      [project @ 2000-11-15 15:43:30 by sewardj] · d68df63f
      sewardj authored
      Overhaul of CM, reducing the complexity of recursive module machinery.
      Also, don't compute package dependencies at all, and don't pass them
      to the linker.
    • simonpj's avatar
      [project @ 2000-11-15 14:37:08 by simonpj] · 8b653a82
      simonpj authored
      The main thing in this commit is to change StgAlts so that
      it carries a TyCon, and not a Type.  Furthermore, the TyCon
      is derived from the alternatives, so it should have its
      constructors etc, even if there's a module loop involved, so that
      some versions of the TyCon don't have the constructors visible.
      There's a comment in StgSyn.lhs, with the type decl for StgAlts
      Also: a start on hscExpr in HscMain.
    • sewardj's avatar
      [project @ 2000-11-15 10:49:53 by sewardj] · 4631557d
      sewardj authored
      Fix up the Batch vs Interactive plumbing.
    • sewardj's avatar
      [project @ 2000-11-15 09:58:00 by sewardj] · 98bae87d
      sewardj authored
      Fix z-encoding problem in findDependency (moduleNameString ->
  6. 14 Nov, 2000 7 commits