This project is mirrored from Pull mirroring updated .
  1. 03 Mar, 2013 2 commits
  2. 01 Mar, 2013 6 commits
  3. 28 Feb, 2013 1 commit
  4. 26 Feb, 2013 1 commit
  5. 25 Feb, 2013 5 commits
  6. 24 Feb, 2013 3 commits
  7. 21 Feb, 2013 2 commits
  8. 20 Feb, 2013 2 commits
    • ttuegel's avatar
      Don't pretty-print unparseable test suite types. · ab9fe920
      ttuegel authored
      Issue #1202. The pretty printer was printing 'type: -' for empty test
      suite interfaces.  The empty test interface type is the result of
      parsing a branch in a conditional block where no type is explicitly
      specified (i.e., the type is inherited from a higher level or propagated
      up from a lower one).  Therefore, the pretty-printer should not print
      any 'type' field at all for test suites with an empty test interface
    • Ben Millwood's avatar
      Update references to bugtracker · f7a1ade7
      Ben Millwood authored
      To reflect the move from hackage trac to github issues. Note that the
      issue numbers don't quite line up.
  9. 19 Feb, 2013 2 commits
  10. 18 Feb, 2013 3 commits
  11. 17 Feb, 2013 3 commits
  12. 13 Feb, 2013 1 commit
  13. 12 Feb, 2013 8 commits
  14. 11 Feb, 2013 1 commit
    • Joachim Breitner's avatar
      Much faster Distribution.Client.Dependency.Modular.PSQ.splits · 895487aa
      Joachim Breitner authored
      This reduces the runtime of "cabal install --dry-run" on a large cabal
      file (550 dependencies) from ~12s to 2.1s. Equivalency with previous
      implementation verified with QuickCheck in ghci.
      This is a guess: The speedup is mostly due to how to the previous
      entries are added to the result of the recursive call. Previously, for
      every entry to be added, the whole result list is mapped over (lots of
      allocations), and then each tuple is taken apart and reconstructed with
      a changed right value (again, lots of allocations). The new code
      assembles a function in an accumulator that represents the required
      update and applies it where needed, hence building the list and tuples
      only once.