Commits on May 17, 2013
  1. Merge pull request #2675 from ulion/curl_range_rework

    Fixed: do not send Range request header when encounter error.
    ulion committed May 17, 2013
  2. CGUIWindowSettingsCategory: update delayed setting changes on focus c…

    …hange asynchronously
    When using the mouse to focus another control after having changed the value
    of a delayed setting that reloads the skin, a SIGSEGV occured because the
    CGUIControl event handler tried to execute code on an already destroyed object.
    Handling the update of the delayed setting change asynchronously allows the
    active event handler to finish working before handling the previously triggered
    GUI_MSG_UPDATE_ITEM which may lead to the destruction of the setting controls.
    Montellese committed May 17, 2013
  3. Merge pull request #2759 from alcoheca/upnp-fix-root-access

    Upnp fix root access
    alcoheca committed May 17, 2013
Commits on May 16, 2013
  1. Merge pull request #2755 from popcornmix/hang_eos_seek

    [rbp/omxplayer] Fix for hang following seek after eos
    huceke committed May 16, 2013
  2. build: don't guess where a feature may be available

    Test the feature, not the OS. Fixes Android build.
    Cory Fields committed May 16, 2013
Commits on May 15, 2013
  1. CGUIWindowSettingsCategory: only handle delayed settings during GUI_M…

    …SG_FOCUSED if the focused setting is not the delayed setting
    When using a mouse to control XBMC changing the value of a spinner with a left
    click results in a GUI_MSG_FOCUSED message. If that setting is delayed a left
    click lead to the early evaluation of the delayed setting even though it's
    still being changed (which should just result in a reset of the delay timer).
    This lead to crashes if a setting changed callback executed a skin reload.
    Montellese committed May 15, 2013
  2. Merge pull request #2753 from Voyager1/libdvdread-fix

    Better Libdvdread fix for SMB/NFS VIDEO_TS dvd files (non-Windows and now including Android) with special thanks to @theuni !
    Voyager1 committed May 15, 2013
  3. [libdvdread] add missing stat() filesystem wrapper - fixes reading VI…

    …DEO_TS dvds on Android
    Voyager1 committed May 15, 2013
  4. Merge pull request #2065 from sandsmark/master

    Drop played parts of video files from the fs cache.
    Arne Morten Kvarving committed May 15, 2013
  5. Drop played parts of video files from the fs cache.

    Patch by Steinar H. Gunderson:
    This patch fixes a persistent problem I've been seing on my Linux HTPC
    in that if I adjust the volume from the XBMC Android remote, the video
    glitches for a few (4–5) frames. Seemingly this is because loading the
    video file has caused the graphics for the volume control to be swapped
    out (it does not happen if I recently used the volume control).
    One could argue that the right fix for this would be to reduce
    /proc/sys/vm/swappiness, but it might not seem reasonable to expect
    end-users to fiddle with VM parameters (and the user might not even have
    root access). I view this fix as complimentary in nature.
    What we do is to use posix_fadvise() to tell the kernel that previously
    read parts of the video file will no longer be used, causing it to drop
    those from its buffer cache. This makes other things, like in my case,
    the volume control graphics, stay cached. This is, of course, assuming
    you don't have other processes that trash your cache, but as long as
    XBMC is running relatively undisturbed, this shouldn't be a big problem.
    The patch is conservative:
     * We only drop data after we've read 16MB past it. (This counter is
       reset on seeks.)
     * We never drop data from the first 16MB of a file (this makes sure we
       don't drop _anything_ from files <32MB, ie. graphics, configuration
       files, etc.).
     * In order to avoid making tons of extra syscalls, we only drop in 1MB
    Unfortunately Linux as of 3.1.0 has no way of just _hinting_ that we
    will no longer need these parts -- when we issue posix_fadvise(), the
    data is immediately and unconditionally dropped from the buffer cache.
    (If you read the manpage, it might seem otherwise, until you come to the
    point where it tells how these flags are actually interpreted by the
    kernel.) Future kernels might remedy this.
    sandsmark committed with Martin T. H. Sandsmark Jan 12, 2013
Commits on May 14, 2013
  1. Merge pull request #2748 from bobo1on1/32bitmul

    Was tested, works.
    bobo1on1 committed May 14, 2013
  2. rename: MUL32 -> AE_MUL32

    bobo1on1 committed May 14, 2013
  3. fixed: when doing a format conversion from float to 32 bit int, multi…

    …ply with INT32_MAX - 127, since this is the maximum value that can be stored in 32 bit float and int, if INT32_MAX gets converted to float, it gets rounded to INT32_MAX + 1 which can cause wrap around distortion
    bobo1on1 committed May 14, 2013
  4. jni: fix crashes when finding external storage. Regression from jni m…

    Cory Fields committed May 14, 2013
  5. Merge pull request #2741 from Karlson2k/win32_sysinfo_clean

    Small cleanup for CSysInfo
    wsoltys committed May 14, 2013
Commits on May 13, 2013
  1. [settings] - fix fetching the eventserver port as int from xbmchelper…

    … - fixes #14329 i guess
    Memphiz committed May 13, 2013
  2. Merge pull request #2744 from herrnst/fix-disable-webserver

    [fix] fix compilation with --disable-webserver
    Montellese committed May 13, 2013
  3. Merge pull request #2665 from Montellese/legacy_db_paths

    add legacy videodb/musicdb path translation
    Memphiz committed May 13, 2013