Commits on May 19, 2012
Commits on May 18, 2012
Commits on May 10, 2012
  1. Windows: skip GetFileAttributes call when opening a file

    piscisaureus committed May 4, 2012
    It wasn't working, and everything seemed to work fine nonetheless. Removing it just saves a syscall.
Commits on May 5, 2012
  1. sunos: make scale better

    Ira Cooper authored and bnoordhuis committed Apr 27, 2012
    Instead of using one port per watch, use one port for all the watches.
Commits on May 1, 2012
  1. unix: map `EROFS` to `UV_EROFS`

    mmalecki authored and piscisaureus committed Apr 14, 2012
  2. err: handle EIO errors on win/unix

    indutny authored and piscisaureus committed May 1, 2012
Commits on Apr 28, 2012
  1. test-spawn failed to build on linux

    erikdubbelboer authored and piscisaureus committed Apr 28, 2012
    - added missing header file
    - converted c++ style comment to c style comment
Commits on Apr 27, 2012
  1. Test for setuid/setgid

    piscisaureus committed Apr 27, 2012
Commits on Apr 24, 2012
  1. linux: add IN_MOVE_SELF to inotify event mask

    bnoordhuis committed Apr 24, 2012
    Partially fixes nodejs/node-v0.x-archive#3172, behavior is now consistent with inotifywait.
Commits on Apr 23, 2012
  1. remove left-over cast fixes #3160

    Igor Zinkovsky
    Igor Zinkovsky committed Apr 23, 2012
Commits on Apr 22, 2012
Commits on Apr 20, 2012

    isaacs committed Apr 20, 2012
Commits on Apr 13, 2012
  1. Change behaviour of uv_flock_destroy and uv_flock_release.

    bacchusrx committed Dec 21, 2011
    Move the unlink call out of uv_flock_destroy and into uv_flock_release (which
    was apparently unused originally), so that uv_flock_destroy only closes any
    open file descriptors and frees memory.
    Then, in uv_pipe_bind, on error, call uv_flock_destroy, but release the lock
    only if the lock isn't held by another process. In uv_pipe_cleanup, do both.
  2. Fix a bug in uv_pipe_bind where the lock file was unlinked prematurely.

    bacchusrx committed Dec 20, 2011
    Added a test to test-pipe-bind-error.c that covers the issue. Basically, since
    uv_flock_destroy gets called whenever uv_pipe_bind fails, only the first
    attempt to bind an already bound pipe would fail. Subsequent attempts would
    For now I've commented it out, but I think some cleanup still needs to happen
    (just not the unlinking when another process already holds the lock).
    I also moved some code out of uv_pipe_bind that was supposed to be in
  3. Reimplement pipe flock that was removed in b2ed24d

    bacchusrx committed Dec 19, 2011
    I appreciate this doesn't work on Solaris (etc?) but it works on Linux
    and Darwin and I need this behaviour.
    After this commit the previously failing test `test-pipe-bind-error.c`
Commits on Apr 5, 2012
  1. add 64bit offset fs functions

    Igor Zinkovsky
    Igor Zinkovsky committed Apr 5, 2012
  2. Windows: backport pipe-connect-to-file fixes from master

    piscisaureus committed Apr 5, 2012
    commit e53ab6675ba12d97ad6d93c9913a473ba5172617
    Author: Bert Belder <>
    Date:   Fri Mar 9 17:04:03 2012 +0100
        Windows: report UV_ENOTSOCK when we opened a file instead of a pipe
        Makes the pipe_connect_to_file test pass on Windows.
    commit 8cbbfbe4c6489868470a7e410f80d4729f4091bf
    Author: Igor Zinkovsky <>
    Date:   Thu Mar 1 14:32:59 2012 -0800
        test: make pipe_connect_to_file succeed with ECONNREFUSED
    commit 6bbccf1
    Author: Igor Zinkovsky <>
    Date:   Thu Mar 1 12:11:12 2012 -0800
        windows: return UV_ENOTSOCK when doing uv_pipe_connect to a file
Commits on Apr 2, 2012
  1. build: define _DARWIN_USE_64_BIT_INODE=1 on OS X

    bnoordhuis committed Apr 2, 2012
    Fixes a segmentation fault on some OS X systems due to sizeof(struct stat)
Commits on Mar 27, 2012