Permalink
Commits on Jul 27, 2011
  1. Hopefully fix nondeterministic crashes on startup.

    swofford committed with vslavik Jul 27, 2011
    The SUUpdateDriverFinishedNotification message must be sent as the very
    last thing in abortUpdate, as it is handled by SUUpdater's
    updateDriverDidFinish, which destroys the driver. That makes subsequent
    sending of Carbon events more than a bit challenging.
    
    Fixes #1.
Commits on Mar 16, 2011
  1. Merge branch 'master' into carbon

    vslavik committed Mar 16, 2011
    Conflicts:
    	SUBasicUpdateDriver.m
Commits on Mar 2, 2011
Commits on Nov 26, 2010
Commits on Oct 14, 2010
  1. Call EVP_MD_CTX_init and EVP_MD_CTX_cleanup.

    Adam Wróbel committed with andymatuschak Oct 14, 2010
    According to openssl documentation EVP_Verify should be accompanied by
    calls to EVP_MD_CTX_init and _cleanup to avoid memory leaks. Not sure, but
    adding for conformity's sake.
  2. Generate keys that actually work with openssl.

    Adam Wróbel committed with andymatuschak Oct 14, 2010
    According to FIPS 186-2 standard DSA keys should be >= 512 and <= 1024 bits.
    Using a 2048 bit key on my system yields an openssl error in Sparkle's
    SUDSAVerifier and causes the update to be rejected.
    
    The error says (can be read only after hacking on Sparkle's code):
    error:0A071066:dsa routines:DSA_do_verify:bad q value
Commits on Sep 13, 2010
  1. Because of the way Sparkle forces SUUpdaters to be per-bundle singlet…

    andymatuschak committed Sep 13, 2010
    …ons, it was possible for an SUUpdater to be dealloced before registerAsObserver was called, which in turn made an exception get thrown on unregisterAsObserver.
Commits on Sep 11, 2010
  1. We used to not register SUUpdater as an observer of the shared user d…

    andymatuschak committed Sep 11, 2010
    …efaults controller if we were running as a bundle. The performance win is minimal, and if I stop doing that, we won't accidentally try to unregister as an observer and bother developers who break on [NSException raise].
Commits on Aug 24, 2010
  1. Fixed delta updates bug wherein updates fail on empty file.

    aidansteele committed with andymatuschak Aug 23, 2010
    mmap()ing empty files caused EINVAL to be returned and _hashOfFile
    would return an incorrect result. This has been rectified.
Commits on Jul 31, 2010
  1. [Bug 611564] [NEW] sparkle fails to build with clang due to -Werror.\…

    andymatuschak committed Jul 31, 2010
    …n\nAlso, switching to Clang because hoorah.
Commits on Jul 28, 2010
Commits on Jul 23, 2010
Commits on Jul 9, 2010
Commits on Jul 5, 2010
Commits on Jun 21, 2010
Commits on Jun 20, 2010
  1. Don't compress in bsdiff.

    vslavik committed with andymatuschak Jun 1, 2010
    bsdiff-produced patch is included in xar archive, where it's compressed
    with bzip2 a second time. Leaving the bsdiff output uncompressed makes
    it possible to compress it better in xar (e.g. using LZMA2 or gzip,
    either of which performs better on some files).
    
    Changes header signature of bsdiff patch file to BSDIFN40 to clearly
    mark the format as different.
    
    Backward compatibility is preserved, bspatch can read both BSDIFF40 and
    BSDIFN40 formats.
Commits on Jun 9, 2010
Commits on Jun 8, 2010
Commits on Jun 6, 2010
  1. Fix 10.4 compatiblity broken by delta updates.

    vslavik committed with andymatuschak Jun 6, 2010
    Delta updates code uses libxar, which is only available in 10.5+, and
    Sparkle framework fails to load on 10.4 because of it. Instead of having
    hard dependency on libxar, Sparkle now links with it weakly and checks
    for its availability at runtime.  This required bumping Xcode project
    compatibility to 3.2 to get weak linking support.
    
    With this change, delta updates are still not supported on 10.4, but
    Sparkle runs there again and just ignores the deltas. This should be
    good enough given 10.4's age and shrinking user base.
Commits on May 4, 2010
Commits on Apr 28, 2010
  1. Fixing ridiculous error in the last commit caused by non-initializati…

    andymatuschak committed Apr 28, 2010
    …on of a variable. See bug #571486. Sorry about that, all.
  2. Added an extra null check to SUAppcast downloadDidFinish because NSXM…

    andymatuschak committed Apr 28, 2010
    …LDocument and NSFileManager throw exceptions on nil arguments. See bug 456514.
Commits on Apr 26, 2010
Commits on Apr 9, 2010
Commits on Mar 29, 2010
  1. Update Carbon port for new initialization code.

    vslavik committed Mar 29, 2010
    Sparkle no longer listens on NSApplicationDidFinishLaunchingNotification
    and so sending it not only doesn't have the desired effect, it causes
    runtime error. Fortunately, initialization is even simpler now, it's
    enough to call init.
  2. Compilation fixes for -Werror.

    vslavik committed Mar 29, 2010
  3. Correct __LP64__ checks.

    vslavik committed Mar 29, 2010
    The macro should be checked for existence, not value.
  4. Initial import of Sparkle for Carbon changes.

    Jack Small committed with vslavik Mar 29, 2010
    Latest known release from
    http://sparkle.downtowncode.com/SparkleForCarbon.html is 1.5b6; this
    import corresponds to it exactly. The code appears to be unmaintained
    and has a crashing bug in it.