Permalink
Commits on Apr 4, 2011
  1. Readme update.

    samcday committed Apr 4, 2011
Commits on Apr 3, 2011
  1. Let's call this commit 0.1.0

    samcday committed Apr 3, 2011
  2. Merge branch 'devel'

    samcday committed Apr 3, 2011
  3. Bump version to 0.1.0

    samcday committed Apr 3, 2011
  4. Added configure option for gitteh, --use-bundled-libgit2, which will

    use the git submodule/included libgit2 in NPM. Forcing npm install to
    use bundled libgit2, since it's updating so often.
    samcday committed Apr 3, 2011
  5. Cleaned up examples.

    samcday committed Apr 3, 2011
  6. Cleanup.

    samcday committed Apr 3, 2011
  7. Documentation done.

    samcday committed Apr 3, 2011
  8. More documentation.

    samcday committed Apr 3, 2011
Commits on Apr 2, 2011
  1. Begun work on documentation.

    samcday committed Apr 2, 2011
  2. Make sure resolve async doesn't release lock on refs until it's loaded

    data for newly resolved ref.
    samcday committed Apr 2, 2011
  3. Updated TODO

    samcday committed Apr 2, 2011
  4. A bit of housekeeping.

    samcday committed Apr 2, 2011
  5. Updated TODO.

    samcday committed Apr 2, 2011
  6. Retard bug fixed.

    samcday committed Apr 2, 2011
  7. Ref packing issues sorted. Seems fairly stable now. Wrote a stability

    test for ref packing. Still needs to be expanded a bit to test for
    situations where a ref is in the middle of deletion for example.
    samcday committed Apr 2, 2011
Commits on Apr 1, 2011
  1. Tree, Ref, Blob, Tag, Commit all use ObjectCache now.

    Also did some cleanups all over the place.
    samcday committed Apr 1, 2011
Commits on Mar 31, 2011
  1. Working on ObjectCache.

    samcday committed Mar 31, 2011
Commits on Mar 30, 2011
  1. Ref packing support. Complete with a crazy bloated implementation

    using 2 mutexes. With an exception, I think I have a ref packing
    solution that will ensure existing loaded references, even if being
    used simultaneously to the packing process, synchronous or
    asynchronous, deleting/deleted or not, will still work.
    
    The exception? I'm getting a race condition using one of the mutexes, 
    I need to come up with a workaround, as I can't free the lock until
    the object has been created, but I can't create the object until I'm
    back on the main loop, which isn't feasible.
    
    I'm worried I may have to overhaul the objectwrapping system a bit, so
    I create the wrapping object anytime I want, and then wrap it in a V8
    JS object when I hit the main loop. I actually just thought of that
    while I was writing this git commit comment. The more I think about
    it, that may be the solution, and it is a bit of an optimization too,
    as I'll be able to load a git object, create a wrapping object for it,
    load all the data, and then wrap it in a JS object when I come back to
    main thread. That would eliminate the need to go into the EIO
    threadpool a second time too, as well as meaning I can allocate the
    C++ wrapping object in the EIO thread (also a minor improvement).
    
    Will have to work on this tomorrow. Bedtime.
    samcday committed Mar 30, 2011
  2. Updated TODO again.

    samcday committed Mar 30, 2011
  3. Updated TODO.

    samcday committed Mar 30, 2011