1. 21 Dec, 2004 3 commits
  2. 20 Dec, 2004 1 commit
    • simonpj's avatar
      [project @ 2004-12-20 17:16:24 by simonpj] · c45a0ac5
      simonpj authored
      --------------------------------
      	Deal properly with dual-renaming
      	--------------------------------
      
      When comparing types and terms, and during matching, we are faced
      with 
      	\x.e1	~   \y.e2
      
      There are many pitfalls here, and GHC has never done the job properly.
      Now, at last it does, using a new abstraction VarEnv.RnEnv2.  See
      comments there for how it works.
      
      There are lots of consequential changes to use the new stuff, especially
      in 
      	types/Type (type comparison), 
      	types/Unify (matching on types)
      	coreSyn/CoreUtils (equality on expressions), 
      	specialise/Rules (matching).
      
      I'm not 100% certain of that I've covered all the bases, so let me
      know if something unexpected happens after you update.  Maybe wait until
      a nightly build has worked ok first!
      c45a0ac5
  3. 18 Dec, 2004 1 commit
    • panne's avatar
      [project @ 2004-12-18 15:45:56 by panne] · f207c9b9
      panne authored
      The documentation breakage continues... >:-( Fixed DocBook. Again: "make html"
      (or the faster "make validate" in the case of DocBook XML) is your friend! >:-(
      f207c9b9
  4. 17 Dec, 2004 5 commits
  5. 16 Dec, 2004 3 commits
    • simonmar's avatar
      [project @ 2004-12-16 10:36:57 by simonmar] · c69886a1
      simonmar authored
      Sanity check when configuring for mingw32: make sure that $CC is a
      mingw gcc before proceeding.
      c69886a1
    • simonmar's avatar
      [project @ 2004-12-16 10:23:44 by simonmar] · 7fd32ef4
      simonmar authored
      --with-gcc: export the CC environment variable, so the setting gets
      picked up by sub-configures.  Perhaps we should be advising people to use
      
        CC=c:/mingw/bin/gcc ./configure
      
      because that works for all configure scripts, not just the top level one.
      
      Background:
      
       - We want --with-gcc to do the right thing, because we don't want
         to require Cygwin users to put c:/mingw/bin first on their
         path: that would break their Cygwin environment.
      
       - The build system should work with *no* gcc on your PATH, as long
         as you use --with-gcc (equivalently, CC=... ./configure).
      7fd32ef4
    • simonpj's avatar
      [project @ 2004-12-16 08:56:15 by simonpj] · 13f7d710
      simonpj authored
      Darn!  Forgot this commit, which killed the nightly build
      13f7d710
  6. 15 Dec, 2004 4 commits
  7. 13 Dec, 2004 1 commit
  8. 10 Dec, 2004 1 commit
    • simonmar's avatar
      [project @ 2004-12-10 13:34:38 by simonmar] · ea453574
      simonmar authored
      stg_sel_##offset##_upd_entry: the ENTER() here is unnecessary, because
      we know the closure in R1 will evaluate to a constructor, so we can
      save some time by just entering it directly.
      ea453574
  9. 08 Dec, 2004 6 commits
    • simonpj's avatar
      [project @ 2004-12-08 15:09:14 by simonpj] · 0e6454e6
      simonpj authored
      Undo bogus 1.4 commit
      0e6454e6
    • simonpj's avatar
      [project @ 2004-12-08 15:03:50 by simonpj] · 3599ea7a
      simonpj authored
      First commit for new Cmm code generation (branch)
      3599ea7a
    • simonpj's avatar
      [project @ 2004-12-08 14:32:28 by simonpj] · 7504acab
      simonpj authored
      Comments and asserts only
      7504acab
    • simonmar's avatar
      [project @ 2004-12-08 12:05:51 by simonmar] · c418e72f
      simonmar authored
      Fix bug #1073501: checkProddableBlock: invalid fixup in runtime linker
      
      The bug manifested when trying to load an object with debugging info
      (compiled with gcc -g) into GHCi.
      
      The problem was that the object loader was ignoring the sections
      containing debugging info, but then it was later trying to do
      relocations for those sections, and its own sanity checking code
      correctly detected that the relocations were in unknown parts of the
      object file.
      
      The fix is to ignore relocations whose target section isn't one of the
      sections that we're interested in, using the same test in both cases
      (the code to test section kind has been extracted).  The code could
      probably benefit from more refactoring: it looks like the list of
      sections we build up in the first phase isn't even used in the second
      phase, instead we traverse the section table in the image again.  This
      looks like cruft leftover from when the GC used to check whether an
      address was in text or data space.
      c418e72f
    • simonmar's avatar
      [project @ 2004-12-08 10:29:01 by simonmar] · 8d750a5c
      simonmar authored
      deleteThread: don't attempt to delete threads blocked in foreign calls.
      8d750a5c
    • simonmar's avatar
      [project @ 2004-12-08 09:28:57 by simonmar] · 5ece0e43
      simonmar authored
      Prevent any intermediate file deletion using
      
        .SECONDARY:
      
      and remove now-unnecessary .PRECIOUS and .SECONDARY targets.
      Suggestion from Ashley Yakeley.
      5ece0e43
  10. 06 Dec, 2004 3 commits
    • simonpj's avatar
      [project @ 2004-12-06 10:58:06 by simonpj] · bfb87645
      simonpj authored
      ---------------------
      	Bug in specialisation
      	---------------------
      
      Laszlo managed to get a function like this:
      
      	foo :: Enum a => (# a, Int #)
      
      The specialiser specialised it, resulting in an unboxed tuple
      binding, which Lint objected to.
      
      This commit adds a dummy argument to the specialised function, 
      very like the case for strictness analysis.  For example, at
      type Char we'd get
      
      	foo_char :: State# RealWorld -> (# Char, Int #)
       	foo_char = \_ -> ...
      
      We use a State# type because it generates no argument-passing code 
      at runtime.  (We should really have some other void type for this 
      purpose, because State# is misleading, but this way avoids extra
      types.)
      bfb87645
    • simonpj's avatar
      [project @ 2004-12-06 10:51:36 by simonpj] · f6f3819f
      simonpj authored
      ------------------------------------
      	Bug in loop detection in TcSimplify
      	------------------------------------
      
      The type-class context simplifier has been able to 
      build recursive dictionaries for some time: co-induction.
      That is, you can build a proof for constraint C by assuming
      that C holds when proving the preconditions of C.
      
      You need to be in -fallow-undecidable-instances land to 
      make use of this: see comments with [RECURSIVE DICTIONARIES]
      in TcSimplify.lhs.
      
      Anyway, this is all fine, but I'd implemented it wrong!  You need
      to be very careful with superclasses, or you can make a bogus
      loop by mistake.  This commit fixes it; tests LoopOfTheDay{1,2,3}
      will test it (thanks Ralf Laemmel).
      f6f3819f
    • simonmar's avatar
      [project @ 2004-12-06 10:47:22 by simonmar] · 23843129
      simonmar authored
      Error message fixup
      23843129
  11. 04 Dec, 2004 3 commits
  12. 03 Dec, 2004 8 commits
  13. 02 Dec, 2004 1 commit