1. 09 Nov, 2015 1 commit
  2. 03 Nov, 2015 1 commit
  3. 02 Nov, 2015 1 commit
  4. 26 Oct, 2015 4 commits
  5. 19 Oct, 2015 1 commit
  6. 09 Sep, 2015 1 commit
  7. 07 Sep, 2015 3 commits
  8. 18 Aug, 2015 1 commit
    • Uri Lublin's avatar
      uxa: fix compiler warnings (initialize width, height, [xy]Dst_copy) · 2f4a2fe2
      Uri Lublin authored
      Some compiler warnings for example:
      uxa-glyphs.c:1045: warning: ‘width’ may be used uninitialized in this function
      uxa-glyphs.c:1045: warning: ‘height’ may be used uninitialized in this function
      uxa-render.c:1090: warning: ‘xDst_copy’ may be used uninitialized in this function
      uxa-render.c:1090: warning: ‘yDst_copy’ may be used uninitialized in this function
      I think these can never actually happen as the code protects use of
      those uninitialized variables, but I'd not "argue" with the compiler.
  9. 10 Aug, 2015 3 commits
  10. 04 Aug, 2015 2 commits
    • Christophe Fergeau's avatar
      Fix compilation with newer Xorg versions · a184774a
      Christophe Fergeau authored
      Xorg 1.18 stopped exporting some xfont related symbols in its
      headers/shared libraries, which causes QXL to fail to build:
      uxa-damage.c:947:5: error: implicit declaration of function 'QueryGlyphExtents' [-Werror=implicit-function-declaration]
           QueryGlyphExtents(font, charinfo, n, &extents);
      The missing definition can be found in xfont, so this commit addes the
      needed configure.ac checks and includes.
      Note that dixfontstr.h must be included before the xfont headers or this
      will cause compile-time warnings on older Xorg versions (eg 1.17)
    • Christophe Fergeau's avatar
      Use <> for system-includes · 83e00bb5
      Christophe Fergeau authored
      These headers come from /usr/include/xorg/ so it's clearer if they are
      included using <>. While at it, I've reordered them alphabetically.
  11. 22 Jun, 2015 1 commit
    • Javier Celaya's avatar
      Fix drawable mm_time with KMS · e765cc5e
      Javier Celaya authored
      When KMS is enabled, the mm_time value of a QXLDrawable is undefined.
      This results in severe synchronization problems. This patch sets it to
      zero to force the spice-server to use the local clock.
  12. 01 Jun, 2015 1 commit
    • Jeremy White's avatar
      Do not process watches on select() error. · d99213d4
      Jeremy White authored
      This enables a kill of an Xorg process to propagate further.
      Without this, the read masks would be set, and we could end up
      blocking in an accept() call and not exiting from the signal.
  13. 22 May, 2015 2 commits
  14. 30 Apr, 2015 1 commit
    • Adam Jackson's avatar
      Use pci_io_write8 instead of outb · c1e88427
      Adam Jackson authored
      This ensures better portability as some arches (eg aarch64) don't have
      On linux pci_io_* try to open the sysfs map file corresponding to the
      I/O port range, which the kernel translates arch-appropriately.  If
      there is no such file then it'll try to use port instructions if that's
      a thing the architecture has.  So you only need one path, pciaccess
      exists to get the portability right for you.
  15. 07 Apr, 2015 2 commits
  16. 02 Apr, 2015 1 commit
  17. 01 Apr, 2015 2 commits
  18. 30 Mar, 2015 2 commits
    • Marc-André Lureau's avatar
      Remove image cache · e4597f71
      Marc-André Lureau authored
      While looking for leaks, I realized that the image cache looks
      quite suspicious.
      Not only it leaks when qxl_drop_image_cache() is called,
      since all the allocated image_info_t references are lost.
      But it is also useless:  "The name implies it's a cache of some
      sort, but data is only added to it. The only time data is looked up in
      that cache is when it needs to be destroyed, and this is not done
      correctly and causes leaks. Since the cached data is never used and
      since it's leaky, it's better to remove it altogether"
    • Christophe Fergeau's avatar
      Don't leak ARGB cursor data bo · 2bed9bff
      Christophe Fergeau authored
      qxl_load_cursor_argb() owns 2 references on the cursor_bo it creates:
      - one from the call to bo_alloc()
      - the second from a call to bo_output_bo_reloc()
      qxl_garbage_collect() release one of these refs, but the other one is
      never released, so ARGB cursor bos are leaked. This can cause out of
      memory issues, for example when running EL6 anaconda installer on a 2TB
      disk image (see bug https://bugzilla.redhat.com/show_bug.cgi?id=1199355
      This commit release the extra ref right after calling push_cursor().
      This is similar to what is done in qxl_surface_put_image().
  19. 20 Mar, 2015 2 commits
  20. 13 Mar, 2015 2 commits
  21. 08 Mar, 2015 2 commits
  22. 25 Jan, 2015 4 commits