[ANNOUNCE] xorg-server

Daniel Stone daniel at fooishbar.org
Tue Jul 31 22:39:40 PDT 2007

This release is mainly to get the foundations for xorg-server 1.4.0, so
we can get the 7.3 release underway (sorry, it's entirely my fault: I've
been blocking the server release on fixing HAL and the modifiers, which
I've just done now).

It's stable enough for me, but I expect that a .1 will come in short
order (within a week, ideally).  In particular, there are a number of
patches on the list that I need to vacuum up and merge.  Some got in,
some didn't.  This release series is _not_ ABI-stable: the input ABI in
particular will break at least once from now until 1.4.0.

Build and run tested on Linux/i386 only.  Testing (and patches) from
other platforms very much welcome.

The major bit this adds is HAL.  If you install config/x11-input.fdi to
/etc/hal/fdi/information, remove your InputDevice sections from
xorg.conf, and add Option "AllowEmptyInput" to the ServerFlags.  This
will have input devices automatically added (and removed) through HAL,
with the appropriate events sent to clients, so that they can manage it.

There's also been a great deal of RandR 1.2 work, the merge of XDarwin,
quite a bit of work on EXA and pixman -- all this and more could be
yours today!

Due to release replanning, this is a direct descendant of
Building it requires Mesa from git (there's no release that builds with
it, and I have zero insight into Mesa's release process, such as it is),
and it also requires pixman.

Shortlog follows the sums, because it's huge.

git tag: xorg-server-

MD5: d37e509df2f4996c147292774f139019  xorg-server-
SHA1: 6c4b1ae0cbfe5eef862bba61f1907d8c8db50937  xorg-server-

MD5: bdd52edba1ad4c9e0826392c42d8eda7  xorg-server-
SHA1: 9fa3c70a80bfc146df18d8e5ad75c66059c6c273  xorg-server-


Aaron Plattner (23):
      Fix standard VESA modes.
      Add a -showDefaultModulePath option.
      Add a -showDefaultLibPath option.
      Add DIX_CFLAGS to util builds.
      Bug #9219: Return BadMatch when trying to name the backing pixmap of an unrealized window.
      Bug #9219: Use pWin->viewable instead of pWin->realized to catch InputOnly windows too.
      Fix BSF and BSR instructions in the x86 emulator.
      Add an RDTSC implementation to the x86 emulator.
      Return BadMatch if a client tries to clone non-cloneable outputs.
      Don't crash setting a NULL mode with a randr classic DDX. Also remember to update the screen size during modesets.
      Add a screen resize hook to xf86CrtcConfigRec.
      Add a canGrow argument to xf86InitialConfiguration.
      Bump video driver ABI to 2.0 for cw change (commit 6ed08949af4f7ac09170d3d9581e4092b24a84ee).
      Swap RRScreenChangeNotifyEvent dimensions when the screen has one crtc and it's rotated.
      Don't call xf86RandR12TellChanged if it doesn't exist. Add some exports to xf86Rename.h.
      Fix unbalanced fbGetDrawable added in commit 0a9239ec.
      Adjust the screen pixmap's dimensions in xf86RandR12ScreenSetSize.
      Include xf86Rename.h in xf86RandR12.h.
      Add new fb symbols to wfbrename.h.
      Handle tileStride > 1 in fbEvenTile.
      Include picturestr.h in xf86Crtc.h to pick up definition of PictTransform.
      Don't fail compScreenInit if the driver added its own alternate visuals.
      Fix a crash when rotating the screen.

Adam Jackson (79):
      Don't translate monitor gamma to X gamma.
      Record the maximum dot clock of the monitor, and filter by it.
      Record all standard timings from EDID as modes, instead of just the first five.
      Allow hsync and vsync ranges to be overridden independently again.
      Enable DPMS by default.
      Fix up EDID blocks where the max pixclock exceeds the preferred mode clock.
      Bump the default pixel depth to 24, and default bpp to 32.
      Publish the raw EDID block as a property on the root window.
      Expand the default sync ranges to be large enough for 800x600 at 60.
      Load the default module set when no Module section is given in the config.
      Attempt to add the 'mouse' driver in more situations.
      Mark EDID modes as driver modes.  Infer virtual size from driver modes.
      Prefer driver-provided modes when matching name strings to modelines.
      In xf86MatchPciInstances, fail gracefully when there's no PCI device at all.
      Always believe the monitor when it reports a reduced-blanking mode.
      Bug #6786: Use separate defines for server's Fixes support level.
      'make dist' fixes.
      Disable Xprt server build by default.
      Accept EDID > 1.3 but < 2.0 if we find it, assume it's compatible.
      Remove spurious LIBADD from xf4bpp
      Hook up --with-builderstring for vendor build identification.
      Typo fix.
      Don't install libi2c.a
      Static markup and dead code cull over xkb/.
      Static cleanups, dead code deletion.
      Delete a dead file.
      Static and dead code cleanup over afb/
      Un-staticise VTSwitchEnabled, since kbd wants it apparently.
      Static cleanup for xf86 ddx.
      Static and dead code cleanup from mi/
      Static and dead code cleaup for Xext/
      General DIX static and dead code cleanup.
      Static cleanup on Xi/
      Since ddc, i2c, and ramdac are in core now, remove their ModuleData stubs.
      Refuse to create tiny modes from EDID detailed timing.
      Move the XAA private indices to be static.
      Don't write out empty sections from the parser.
      VT activate or waitactive are fatal if they fail.
      Look for the PCI ROM file elsewhere in sysfs.
      Fix a buffer overrun on machines with excessively large PCI busses.
      Disable RANDR's fake Xinerama protocol when there's more than one screen.
      Paper over a crash at exit during GLX teardown.
      Remove old edid_modes.c, it lives in hw/xfree86/modes/ now.
      Fix the 'relink' target for kdrive servers.
      Make the use of ICEIL slightly less ugly.
      Use _X_INLINE instead of ad-hoc #defines.
      Use system copy of cbrt() if available.
      Remove mfb and cfb from include paths where they're not needed.
      Bug #8991: Add glXGetDrawableAttributes dispatch; fix texture format therein.
      Delete VDIF support; it was never used anyway.
      Don't print lack of DRI support as an error in AIGLX init.
      Don't warn about default behaviour when autoconfigging.
      Minor cleanup/robustification to config parsing.
      Remove the old Kerberos 5 authentication code.
      GNU is wrong and ` is not left-quote.
      Make xf86{En,Dis}ableInterrupts no-ops on Linux.
      Add VBE PanelID support.
      Bug #10770: "Inputdevs" isn't a valid config file keyword.
      Fix another usage of MAX_PCI_DEVICES.
      Remove the remnants of OS/2 support.
      Death to RCS tags.
      Make x86emu's I/O cycle tracing more useful.
      Clean up unused #ifdefs from fb.
      Add per-monitor config file option for maximum pixel clock.
      Nuke dead X -configure code.
      Dead ifdefs for BITMAP_SCANLINE_UNIT == 64
      Remove MEMDEBUG
      Delete dead module test code.
      Remove dead code for screen crossing.
      Always normalize the module name.
      Remove (long-)deprecated xf86EnablePciBusMaster.
      Delete some pre-dlloader debugging scaffolding.
      Refactor how Composite adds visuals to the screen.
      Partial redundancy elimination in PropertyNotify generation.
      libconfig shouldn't be an installed library.
      exaGetPixmapFirstPixel: avoid framebuffer readbacks if possible.
      Remove all trace of Option "BiosLocation".
      ReduceCompositeOp returns a Render op, not a boolean.

Alan Coopersmith (37):
      PostSyntheticMotion needs to be extern, not static, since it's in getevents.c
      sun_bell.c needs to #include "xf86_OSlib.h"
      Remove references to xf86Info.kbd* from solaris code
      Convert Xprt DDX to new motion history api
      Add DTrace probe points for X server <-> client communications
      Correct symlink-mesa.sh usage message
      Tell automake to STFU about the *.O files
      Make solaris version of xf86OSRingBell return void like other OS'es
      Fix syntax error in configure check for SYSV_IPC that broke with Sun cc
      Check for __sparc as well as __sparc__ for compatibility with Sun cc
      Update pci.ids to 2006-12-06 from pciids.sf.net
      Xorg & Xserver man page updates for 1.2 release
      xorg.conf man page should say "XFree86-DGA", not "Xorg-DGA"
      Make GLX byteswap macros more portable
      Use PKG_CHECK_EXISTS(libdrm) to determine if DRI should be enabled on Solaris
      Always include compiler.h in cfbmskbits.h instead of checking #ifdef XFREE86
      Xserver man page: remove bc, add -wr
      Make xf1bpp build correctly with compilers that don't support -include
      Update pci.ids to 2007-01-18 snapshot
      Update Xserver man page to match commit ed33c7c98ad0c542e9e2dd6caa3f84879c21dd61
      Correct variable descriptions in comment for SecurityCheckResourceIDAccess
      Correct help lines for configure's --with-vendor-name flags
      Plug memory leak in doLoadModule()
      Fix bus error on startup in 64-bit Xephyr
      Update Xvfb man page: remove monolith build instructions, use /var/tmp instead of /usr/tmp
      Move SecurityPolicy file format from Xserver(1) to it's own man page
      Sun bug 6529003: Xorg should not be including <sys/immu.h> on Solaris
      Define XF86PM on Solaris x86 builds now that we have sun_apm.c
      Include module name in "already built-in" message
      Split checks for dtrace & getpeerucred()
      "fbpict.c", line 215: void function cannot return value
      Make SOLARIS_INOUT_ARCH substitutions work better with automake-1.10
      Use kbd driver when xorg.conf specifies "keyboard" or "Keyboard" (bug #11301)
      Add __SOL8__ to xorg-server.h.in since xf86-input-kbd needs it to build
      Use %S instead of %s for strftime seconds when printing build time
      Update pci.ids to 2007-07-16 snapshot
      Include comment/copyright/license for AC_DEFINE_DIR in acinclude.m4

Alan Hourihane (12):
      Shutdown input devices if FatalError occurs during startup.
      Issue CloseDownDevices() in os/log.c and remove from dix/main.c.
      Fix Tooltip from minimized clients
      Fix Xming fails to use xkb bug
      Fix bad commit
      Set Int10Current->Tag for the linux native int10 module
      lnx_ev56.c has to be built with -mcpu=ev56. Fix it.
      Remove array_cache for recent Mesa changes
      remove file
      clean up more of the vbo fallout
      Set pScreen on context
      Fix mode validation against the maximum X/Y values configured

Alberto Mardegan (1):
      Bug #6620: Fixed a missing 'else' in ATIPseudoDMAInit().

Arcady Goldmints-Orlov (1):
      Fixed fbSolidBoxClipped() to fill the right place.

Arkadiusz Miskiewicz (1):
      XFree86: xf1bpp: Fix parallel build

Bastian Blank (1):
      Fixes for s390

Ben Byer (33):
      Beginning of patches to add XDarwin support to the modular tree;
      iokit support for XDarwin
      Beginnings of an update Darwin driver
      merged in miext/rootless changes for XDarwin support
      updated darwin/quartz/xpr (libXplugin interface for Mac OS X) support
      More build fixes / updates for XDarwin:
      more patches to make the Quartz part of XDarwin work again
      glx fixes for XDarwin
      build fix for configure.ac / BUILD_DARWIN, oops
      dix mods for Darwin
      autoconf goodness for XDarwin, courtesy of pogma
      cleaned up some linking ugliness in hw/darwin/quartz
      oops, missed a spot
      configure fixes for darwin
      added hw/darwin/apple directory, which contains source and data files to build
      more changes for X11.app
      X11.app now builds correctly
      added hw/darwin/README.apple file with some todo items and props.
      verbiage corrected per daniels
      fix for hw/darwin/Makefile.am to properly use XSERVER_LIBS
      AIGLX support for Darwin -- works well enough to run
      Rewrote parts of the Xquartz event-handling code (thanks daniels and whot!)
      Fixed pointer events in Xquartz -- Keyboard events work, but
      autoconf fixes for XDarwin (created DARWIN_LIBS)
      fixed X11.xcodeproj to get CFLAGS and LDFLAGS from autoconf script
      Makefile fix for X11.app
      stopped using XTrans internals in X11.app because they're apparently no longer public
      started moving new input code into darwinEvents.c so that it may be shared by the three servers
      began to factor out code to move to darwinEvents.c
      Fixed Darwin's Makefile.am to fix a problem building X11.app
      moved new event-handling code from X11Application.m to darwinEvents.c in preparation for making all Darwin servers use it
      rewrote event handling, Xquartz now has working mouse and keyboard. use it\!
      updated todo list

Benjamin Herrenschmidt (1):
      Include pixman.h from fb.h or compile of some files will fail

Bernhard Rosenkraenzer (1):
      automake: avoid use of reserved _SOURCES keyword (bug #8866)

Bjorn Helgaas (2):
      Do not map full 0-1MB legacy range
      Bug 9041: Check the return code in xf86MapDomainMemory().

Bram Verweij (1):
      xfree86/linux acpi: fix tokenising

Branden Robinson (1):
      Overhaul xorg.conf manpage

Brian (36):
      Regenerated from Mesa, fixes glMap* protocol problem (bug 8899)
      Regenerated from Mesa w/ latest gl_API.xml file
      ompile fbcmap.c w/ -DXFree86Server instead of linking libfbcmap.a.
      Fix some bad formatting.
      more formatting fixes
      Checkpoint fixes to DMX for X input changes.
      more debug
      num_valuators=1 for GetPointerEvents(), hack ButtonPress/Release position
      Tweak some parameters, etc.  Things seem a little better now, but still a ways to go.
      Checkpoint DMX updates: things are working much better now, but still not 100% right.
      formatting fixes
      formatting fixes
      Pass num_valuators=0 for ButtonPress/Release. This seems to fix the button coordinate problem
      undo 1280 valuator hack
      In dmxBackendMouGetInfo() initialize the info->minval[], maxval[] arrays to the size of the backend display.
      As for normal mouse device, init valuator maxval[] to real values, not zero.
      for completeness, init dummy's min/maxval[1] values (vertical axis)
      remove some debug code
      checkpoint clean-up
      clean-up dmxCoreMotion() dmxCoreMotion2()
      formatting fixes
      checkpoint: more clean-up
      fix formatting
      clean-up, debug code
      Split the xserver/fb/fbcmap.c file into two files.
      add DDXRingBell() stub to solve link problem
      Implement a minor hack in dmxCheckFunctionKeys() to detect special keys.
      regenerated to fix bug 10371
      remove sources deleted in Mesa
      regenerated to add GL_CLIENT_ATTRIB_STACK_DEPTH (bug 9823)
      replace occlude.c w/ queryobj.c
      add slang_mem.c
      fix bug in which maxKeysPerModifier wasn't getting set
      bump release date to reflect input code updates
      In __glXCreateARGBConfig(), insert the new GL mode at the _end_ of the linked list.

Brice Goglin (2):
      Bug #10034: 'man Xserver' typos: s/dqoute/dquote/
      Minor fixes in cvt and gtf manpages

Clark Rawlins (1):
      Really make sure BUILD_TIME doesn't have a leading zero.

Colin Guthrie (1):
      fix __glXErrorCallBack() proto

Colin Harrison (4):
      Migrate some code to the new mi apis
      Fix bad use of hwnd (bug: 9808)
      fix an occasional crash in GetWindowName() (bug: 9798)
      Missing piece from bug 9808

Dan Nicholson (1):
      The array_cache sources don't exist anymore in the Mesa tree,

Daniel Ciocea (1):
      Fix sync polarity on Samsung SyncMaster 205BW monitor.

Daniel Drake (3):
      xserver: Add COPYING terms
      Add proper COPYING file
      Remove duplicated licenses

Daniel Stone (87):
      RemoveGeneralSocket: don't touch EnabledDevices
      mieq: annotate with some more comments
      xfree86/os-support: update prototype for OSRingBell
      xfree86/xf86misc: warning fixes
      dix: remove unused variable
      dix: remove unused debug code
      dix: add missing prototypes
      dix/mi: still more warning fixes
      xace: avoid 'unused variable pScreen'
      os: fix sun extensions test
      xkb: fix uninitialised warning
      XkbCopyKeymap: don't iterate broken types, or dereference null pointers
      dix: remove staggeringly broken vendor workarounds
      SyntheticMotion: don't dereference sprite.screen when not using Xinerama
      xkb: warning fix
      config: error messages non-fatal, and before we free them
      kdrive/sdl: fix compilation
      configure.ac: axe redundant SDL test
      Get*Events: massive reorganisation
      xfree86: add general handler, port ACPI to it (bug #5665)
      xfree86: remove xf86Keymap.h
      remove trailing whitespace
      remove CID support (bug #5553)
      SyntheticMotion: don't dereference sprite.screen when not using Xinerama
      xfree86/input: re-add support for disabling drag events
      kdrive/tslib: remove vendor-specific hacks
      config: move to block/wakeup handler
      kdrive/mga: fix compiler warning
      configure.ac: add CONFIG_LIB to Xvfb
      GetPointerEvents: fix typo
      GetPointerEvents: always send valuator events for MotionNotify
      config: move config.h to hotplug.h
      config: bus reconnect support
      whitespace police
      xfree86: remove stray debug line
      XkbCopyKeymap: always initialise map and preserve
      XkbCopyKeymap: make sure sym_interpret is always valid
      os: test for userland, not kernel
      dmx, vfb, xnest: fix fbcmap compilation
      XkbCopyKeymap: fix copy-and-waste accident
      kdrive/ephyr: free screen struct
      change versioning for new server version scheme
      configure.ac: add xdarwin stubs
      os: fix client privates leak
      kdrive: delete input debugging, yet again ...
      kdrive/ephyr: fix keysym type confusion once and for all
      config: error message cleanup
      configure.ac: disable dmx per default
      mieq: Allow event handlers for arbitrary events to be set
      xfree86 input: Re-enable DGA support
      XFree86 input: Add backwards compatibility for motion history
      XKB: Be a tiny bit more conservative with type allocation
      XKB: Fix size_syms calculation bug
      XFree86 DGA: Guard against NULL pointer dereferences.
      GL: Update for Mesa changes
      XFree86: Treat evdev and vmmouse as mouse drivers (bug #10512, #10559)
      mieq: Use larger default queue size
      getevents: Copy modifier state from extended to core devices
      XFree86: DGA: Don't call ProcessInputEvents from CloseScreen
      XFree86: Input: Perform case-insensitive comparisons on option names
      XFree86: Input: Assume core events per default
      configure.ac: Fix KDrive VESA/fbdev conditionals
      configure.ac: Properly check XFree86 proto modules
      gitignore: Add automake lex/yacc wrapper
      Hotplug: Separate D-Bus into core and hotplug API components
      Hotplug: D-Bus: API version 2
      Fonts: Fix builtin fonts
      dix.h: Remove duplicate ffs() prototype.
      DIX: Clean up null root cursor handling
      Input: Allow enabling and disabling of devices
      Input: GetPointerEvents: Deny events from devices without valuators
      Hotplug: D-Bus: Dispatch harder
      Hotplug: Add HAL support
      Convert all my license statements to the standard form
      XFree86: Module: Bump input version
      XFree86: Input: Fix whitespace
      Config: D-Bus core: Fix hook removal
      Input: Don't enable devices when we open them
      XFree86: Allow disabling of HAL
      Input: Fix stuck modifiers (bug #11683)
      Config: Fix merge detritus
      Config: Add current FDI file
      Config: Add missing include
      Darwin: Remove missing file
      Build system: Non-dtrace distcheck hacks
      XFree86: Properly clean up after ourselves
      Build system: Add missing files

Dave Airlie (13):
      make X server use system libdrm - this requires libdrm >= 2.3.0
      dri: setup libdrm hooks as early as possible.
      update xserver for vbo code in mesa
      add vbo to .gitignore
      remove array_cache from everywhere
      modes: add commit/prepare hooks
      add a standard connector type and name for us as an output property
      loader: fix already built-in message
      fix loading of GLcore after recent loading changes
      gl: update for latest mesa glsl-compiler merge
      gl: oops dodgy s appeared pointed out by jcristau on irc..
      update xproto dependency to at least 7.0.9
      Revert "Fix RandR 1.2 conversion of two colour to ARGB cursor on MSB first platforms."

David Nusinow (7):
      Improve modules loading defaults
      Provide UseDefaultFontPath option
      Add more informative logging for module default loading
      Fix boolean thinko that prevented working without a server layout
      Fix up xnest manpage
      Read ROM in chunks
      Fix for GNU/kFreeBSD

Dodji Seketeli (2):
      ExaOffscreenMarkUsed: Don't crash when there's no offscreen memory.
      exaDriverInit: Fail if pScreenInfo or a member of it is invalid.

Drew Parsons (4):
      GNU/FreeBSD support in GLX: include byteswap.h for FreeBSD systems
      Xprint: fix font symlinks
      Xprint: shorten font filename to fit in tar length limit
      Update Xprint build for pixman.

Eamon Walsh (27):
      Bug #8875: Security extension causes Xorg to core dump on server reset
      whitespace adjust
      More work on Bug #8875: revert previous fix and try using client argument
      Bug #8937: Extension setup functions not called on server resets
      Define calls away when not building XACE, allowing ifdef's to be removed.
      Naming change: Security*Operation -> Xace*Operation
      Naming change: Security*Access -> Dix*Access
      Add new, combined dix lookup functions.
      Add new, combined dix lookup functions (tweak).
      Remove instances of macros LOOKUP_DRAWABLE and VERIFY_DRAWABLE.
      Remove instances of macro SECURITY_VERIFY_DRAWABLE.
      Remove instances of macros VERIFY_GEOMETRABLE and VERIFY_GC.
      Remove instances of macros SECURITY_VERIFY_GEOMETRABLE and SECURITY_VERIFY_GC.
      Remove now-unused macro definitions from dix.h.
      Convert callers of LookupWindow() to dixLookupWindow().
      Convert callers of SecurityLookupWindow() to dixLookupWindow().
      Convert callers of LookupDrawable() to dixLookupDrawable().
      Convert callers of SecurityLookupDrawable() to dixLookupDrawable().
      Convert callers of LookupClient() to dixLookupClient().
      Add loud warnings to deprecated lookup functions.
      Zero out client devPrivates on allocation.
      Bug #6988: Change behavior of Security extension per user feature request.
      Make mfb, cfb, and afb support configurable at build-time.
      remove PIXPRIV checks as this flag is always set.
      remove PIXPRIV check as this flag is always set.
      Properly free device devPrivates - memory leak fix.
      Remove dead NEED_DBE_BUF_BITS code.

Egbert Eich (2):
      Make int10 fully domain aware.
      Fixing mach64 driver bailing out on ia64

Eric Anholt (55):
      Provide option to report damage after operation is complete.
      Don't bump the refcnt if the new mode is NULL.
      Fix build on FreeBSD after input-hotplug.
      Report the error before freeing the error strings.
      A couple more cases of error message before freeing strings.
      Bug #8868: Remove drm from SUBDIRS now that the directory is gone.
      Fix typo before the last commit.
      Fix build on FreeBSD after input-hotplug.
      Add some mode helper functions from the intel driver.
      Move mode handling helpers from ddcProperty.c to xf86Mode.c.
      Register dependency on new RandR protocol.
      Typo that was missed in testing.
      Move code to get a mode list from EDID data from ddcProperty.c to edid_modes.c.
      Clean up a bunch of long lines and trailing whitespace.
      Replace bad mode name-setting code with xf86SetModeDefaultName().
      Separate DDC mode list creation from MonPtr creation.
      EXA: Disable SHM pixmaps.
      Include sys/select.h to get FD_ISSET.
      Only do the _POSIX_C_SOURCE hackery on linux where it's required.
      Export exaMove{In,Out}Pixmap().
      Switch the default migration heuristic for EXA to "always".
      Try dlsym(RTLD_DEFAULT) first when finding symbols.
      Keep track of how many visuals we set up for GLcore, to avoid an invalid free.
      Add support for the DamagePost (XDamage 1.1) request.
      Track rename of DamagePost -> DamageAdd.
      Bug #9555: Always define _GNU_SOURCE in glibc environments.
      Don't forget to add the property we configure to the properties list.
      When changing a non-pending property, call the screen rrOutputSetProperty hook.
      Add a setter for randr_crtc->rotations.
      Account for CRTC rotation in the cursor containment code.
      Really fix optimized render cases being hit when they shouldn't.
      Bug #7639: Only swap out pixmaps (rather than everything) on VT switch in EXA.
      Warning fix for RRCrtcSetRotations().
      Bug #9680: Remove bogus blank length limiting in xf86SetModeCrtc().
      Restore a few important lines killed in the previous commit.
      Set the Damage version supported in the server, instead of using damageproto.
      Fix the size expectations of xRRSetCrtcGamma.
      Mark sync when UploadToScreen succeeds in exaGlyphs().
      Add missing dirty marking in a couple of fallback cases in the exaGlyphs path.
      Bug #9931: Fix linear allocations with a non-1-byte granularity.
      Set the RandR version returned, rather than just passing the proto's version.
      Update xorg.conf manpage for new RandR 1.2 monitor options.
      Fix indentation of fakexa help text.
      Move libcw setup to the only renderer requiring it (XAA).
      Refuse to initialize Composite if Render is not present.
      Disable Composite when the screen's visual is pseudocolor.
      Disable composite when Xinerama is active.
      Enable Composite by default now that it disables itself in the known bad cases.
      Move modes/ debugging output under Option "ModeDebug" in the Device section.
      Remove libminimi build.
      Belatedly bump XORG_VERSION for 7.2.
      Suppress software cursor removal during rotated shadow buffer drawing.
      Fix bswap detection on BSD (mis-added '_' in function names).
      Fix documentation of association of outputs to monitor sections in xorg.conf(5)
      Correct the xf86EdidModes.c file description.

Erik Andren (1):
      remove XFree86 changelogs (bug #7262)

Erik Andrén (1):
      Syncmaster 226 monitor needs 60Hz refresh (#10545).

Eugene Konev (1):
      Allow configurable serverconfigdir for security policy location

Fredrik Höglund (2):
      Move the code for resetting the DPMS mode in response to input events,
      Add a new IDLETIME system sync counter.

George Sapountzis (12):
      DRI: call drmSetServerInfo() before drmOpen().
      Fix GL context destruction with AIGLX.
      exaGlyphs: mark dirty for software path also.
      dmx: drop leftover __GLXdrawablePrivateRec struct.
      Make xf86glx.c unaware of Mesa internals
      glx: fix symlink, glcontextmodes.c was moved to mesa/src/glx/
      glx: Remove stray __GLinterface.
      configure: minor cosmetic, move GLX extension options together.
      GLcore: fix after moving xf86glx_util.[hc] to Mesa.
      glx: drop stray CAPI define.
      glx: drop xmesaP.h include from xf86glx.c
      glx: move __glXMesaProvider from GLcore module to glx module.

Gerhard Tonn (1):
      Miscellaneous fixes for S/390.

Gero Mudersbach (1):
      Bug #10814: Add needed quirk for Samsung 225BW like the 226BW.

Gustavo Pichorim Boiko (2):
      Set the crtc before the output change is notified
      Fix the output->crtc initialization in the old randr setup

Hanno Boeck (1):
      xnest: fix linking since dbus

Henry Zhao (1):
      Need to use minPitch in miScanLineWidth() to get the shrinked

Ian Romanick (5):
      Add new header file containing byte-order wrappers.
      Regenerate from Mesa scripts.
      Incorporate new byte-order related configure changes.
      Re-regenerate from Mesa scripts.
      Add missing #else from previous commits.

James Steven Supancic III (1):
      Fix RENDER issues (bug #7555) and implement RENDER add/remove screen

Jared D. McNeill (1):
      Add a real xf86EnableIO/xf86DisableIO for NetBSD/PPC.

Jay Cotton (1):
      Add sun_apm.c for Suspend-and-Resume support on Solaris

Jay Estabrook (2):
      Fix root bus/domain selection on alpha
      Ensure domain is stripped from the bus ID.

Jens Granseuer (1):
      Bugzilla #7145: fix build with gcc 2.95

Jeremy C. Reed (3):
      On DragonFLy, default to /dev/sysmouse (just like on FreeBSD).
      For NetBSD, define PCVT_SUPPORT (System has PCVT console).
      For MANDEFS, also replace __mandir__ for $(mandir) which includes

Jesse Barnes (1):
      Fix documentation for Copy hook -- it can copy memory to the scanout

Jonathan Lim (2):
      Call linuxPciOpenFile() for r/w access if applicable.
      Bug 5000: Fix domain support for SGI Altix

Julien Cristau (2):
      Make sure that the ramdac symbols are present in the server
      Fix alpha build failures

Juliusz Chroboczek (2):
      Fix typo in Xvesa: incorrect reporting of DAC capabilities.
      Improve vm86 error handling in Xvesa.

Jurij Smakov (1):
      ffs: handle 0 argument (bug #8968)

Keith Packard (108):
      Preliminary RandR 1.2 work
      RandR: New data structure, old API. At least it compiles now
      Successful legacy RandR API/Protocol emulation for query.
      Start moving to new randr 1.2 definition
      Split RandR implementation into separate files.
      Split out RandR dispatch code from randr.c to rr*dispatch.c.
      Remove smashing of CFLAGS from server build.
      RandR working with old clients and old API.
      RRGetScreenResources and RRGetOutputInfo are working now.
      Split out 1.0-style info and new property routines to their own files.
      Avoid calling xalloc(0). Change rrScreenSizeSet to rrScreenSetSize.
      When setting output state, leave output unchanged when setting to current.
      RRSetCrtcConfig status fix. RRGetScreenResources timestamp fix.
      Steal Xinerama code from SiS driver. Add missing files.
      When no mode is specified, don't validate mode-specific parameters.
      Add mode origins and output options. Fix memmoves in resource free funcs.
      Add preferred modes for each output. Round vrefresh. Deliver crtc events.
      Limit pointer to valid crtc areas. Add event swapping. Fix change tracking.
      Merge master back in and clean up some unfinished code (closes 8745)
      Move physical size from mode to output.
      Define fbHasVisualTypes in fb.h as it is exported
      Add DIX_CFLAGS to hw/vfb/Makefile.am
      Allow X server to build against libdrm 2.1
      Avoid dereferencing sprite.screen when Xinerama is not running. (#8925)
      Add $(DIX_CFLAGS) to remaining Makefile.am files
      Allow RandR objects to be created before the associated ScreenRec.
      Add RRInit function to create resource types for RR objects.
      Remove RandR output options.
      Reduce calls to RRGetInfo.
      Change RandR property datatype to include pending/valid values.
      Allocate correct size for RRPropertyRec (oops).
      Ensure RandR resource types are registered before resources are created.
      RandR ListOutputProperties has nAtoms element, not nProperties
      Destroying RandR crtc or output overwrites memory.
      RandR mode list needs both output and crtc modes.
      RandR: config time updates when hardware config changes.
      RandR 1.0 refresh rates unscrambled. SetScreenConfig uses RRCrtcSet right.
      RandR 1.2 rotation code must adjust width/height.
      Use RRScreenSetSizeRange in 1.0 compat. Check RRGetInfo for error.
      Track physical screen size and send out updates when that changes.
      Make Xinearama screen information reflect CRTC rotation.
      Merge crtc/output-based mode selection code.
      Allow new modes code to build inside drivers as well as server.
      Report correct RandR 1.0 sizeID. Report correct subpixel order.
      Kludge around duplicate code added in hw/xfree86/modes.
      RRConfigureOutputProperty is a variable length request.
      Enable startup-time rotation; change rotation pixmap creation API.
      Respect rotation in initial screen size computation.
      Ensure drivers can use new modes header files.
      Don't set subpixel order during startup; the screen won't be ready.
      Check for clientGone before sending events from XFixes (bug #1753).
      Handle non-zero origin rotated crtc. Damage crtc area on re-rotate.
      Remove debugging ErrorF from rotation code.
      Move xf86SetSingleMode into X server from intel driver.
      Add xf86SetDesiredModes to apply desired modes to crtcs.
      Use EDID data to set screen physical size at server startup.
      Allow relative positions to use output names or monitor identifiers.
      Add xf86CrtcScreenInit to share initialization across drivers.
      Add hw/xfree86/docs/README.modes, documenting new mode setting APIs.
      Remove stale monitor data when output becomes disconnected.
      Create driver-independent CRTC-based cursor layer.
      Allow xf86_reload_cursors during server init.
      Don't wedge when rotating more than one CRTC.
      Add support for user-defined modelines in RandR.
      Eliminate RRModeRec devPrivate field.
      Correct ref counting of RRMode structures
      Remove extra (and wrong) I2C ByteTimeout setting in DDC code.
      Slow down DDC I2C bus using a RiseFallTime of 20us for old monitors.
      Clean up Rotate state on server reset.
      Clear allocated RandR screen private structure.
      Clean up xf86CrtcRec and xf86OutputRec objects at CloseScreen.
      Make sure RandR events are delivered from RRCrtcSet.
      Fix Pending property API, adding RRPostPendingProperty.
      Incorrect extra memory copy in RRChangeOutputProperty.
      Ensure that crtc desired values track most recent mode.
      Make pending properties force mode set. And, remove AttachScreen calls.
      Don't erase current crtc for outputs on CloseScreen
      Add setrlimit call in -core option to make dumps occur.
      In AIGLX EnterVT processing, invoke driver EnterVT before resuming glx.
      Disable CRTC when SetSingleMode has no matching mode. Update RandR as well.
      Rotate screen size as needed from RandR 1.1 change requests.
      Add quirk for Acer AL1706 monitor to force 60hz refresh.
      RandR 1.2 spec says CRTC info contains screen-relative geometry.
      Using wrong log level in extension to built-in message
      Use default screen monitor for one of the outputs.
      Allow outputs to be explicitly enabled in config, overriding detect.
      Revert "Suppress software cursor removal during rotated shadow buffer drawing."
      Was accidentally disabling rotation updates in mode set.
      Disable SourceValidate in rotation to capture cursor.
      Fix for a divide by zero that can be triggered by a malicious client.
      Disable all outputs and crtcs at startup.
      Use Screen block handler for rotation to draw under DRI lock.
      xf86XVFillKeyHelper assumed root clip never changed.
      Skip driver mode detection/configuration when !vtSema.
      Make Composite manual redirect windows not clip their parent.
      Force advertised Composite version to 0.4 instead of using header version.
      Have Composite always report server version.
      Fix MEMORY SMASH in XkbCopyKeymap.
      Generate ChangeLog file for make dist.
      Redirect fix: Manual + Automatic - Manual = Automatic
      When sync'ing logfile, also flush it.
      Screen size bounds check in ProcRRSetCrtcConfig not masking out reflections.
      Add RandR reflection support.
      MakeAtom needs length without trailing NUL. sizeof("string") includes NUL.
      Make pending property changes trigger mode setting.
      Query modes on disabled (but not ignored) outputs.
      Make PreferredMode option in config file override EDID mode preferences.

Kevin E Martin (1):
      For Xvfb, Xnest and Xprt, compile fbcmap.c with -DXFree86Server

Kristian Høgsberg (1):
      Don't map the front buffer in libdri if the ddx driver doesn't set the size.

Laurence Withers (1):
      CreateColormap: fix return value (bug #7083)

Lennart Buytenhek (1):
      Compile fixes for Linux ARM platforms.

Luc Verhaegen (1):
      Bug #5386: Synthesize modelines from EDID data.

Luo Jie (4):
      Fix build of composite, dix, and randr when Xinerama is disabled.
      Fix os/utils.c compile with mingw.
      Fix a typo in using memcpy in xwin.
      Reinstate an apparently mis-deleted ';' from a for loop with no body.

Magnus Vigerlöf (6):
      Input: Plug memory leak in device free
      XFree86: Fix memory leak in option parsing
      XFree86: Fix memory leaks, option parsing, in NewInputDeviceRequest
      Input: Add DeleteInputDeviceRequest
      Config: Fix memory leaks
      Input: Always add devices with first available ID

Marc Aurele La France (1):
      xfree86: deal with pitch that isn't a multiple of the granularity

Matthias Drochner (1):
      Fix build on NetBSD/amd64.

Matthias Hopf (13):
      Build with -D_PC on ix86 only.
      Added missing domain stripping in already domain aware code.
      Added linux 2.6 compatible domain aware device scanning code.
      Fix obviously wrong boundary checks + cleanup unused vars.
      Fix 2 warnings.
      Fixing domain support for ia64
      Add domain support to linuxPciOpenFile().
      Fix device path in altixPCI.c to be domain aware.
      Bug #9023: Only check mice for "mouse" or "void" if identifier is != NULL.
      Fix potential NULL pointer access in timer code.
      Legacy framebuffer support wasn't compiled if Xorg wasn't explicitly enabled.
      Fix calculations in x86 emulator for the long long case (Andreas Schwab).
      Disable Simba PCI bridge routing code (Bug #8020).

Matthieu Herrb (5):
      Multiple integer overflows in dbe and render extensions
      CVE-2007-1003: XC-MISC Extension ProcXCMiscGetXIDList() Memory Corruption
      Make this build on OpenBSD
      Remove wscons keyboard handling stuff that doesn't belong there anymore.
      swap xOrigin and yOrigin in SProcRenderSetPictureClipRectangles.

Michel Dänzer (57):
      Track per-drawable damage to minimize UTS and DFS transfers.
      exaDrawableDirty: Fix initialization of BoxRec.
      xfree86 DDX: Delete DDX screens in ddxGiveUp().
      Revert "xfree86 DDX: Delete DDX screens in ddxGiveUp()."
      EXA: Compare backing pixmaps instead of drawables against driver limits.
      exaCopyNtoN: Fix usage of 'dx' and 'dy' instead of 'reverse' and 'upsidedown'.
      EXA: Lots of damage tracking fixes.
      fbdevhw: Consolidate modeset ioctl calling, report failure if it modifies mode.
      fbdevhw: Fix some issues with the previous commit.
      fbdevhw: Use displayWidth for fbdev virtual width when appropriate.
      fbdevhw: Override RGB offsets and masks after setting initial mode.
      Attempt to fix drawable type checks in dixLookupDrawable().
      Require glproto >= 1.4.8 for GLX.
      __glXDRIscreenProbe: Use drmOpen/CloseOnce.
      fbdevhw: Consider mode set equal to mode requested if virtual width is larger.
      fbdevhw: Only deal with RGB weight if default visual is True- or DirectColor.
      Bump video driver ABI version to 1.2.
      Track number of visible DRI windows separately for transitions.
      DRI: New ClipNotify driver hook.
      DRI: Make sure number of DRI windows is accurate in driver ClipNotify hook.
      Fix typo in GL/mesa/shader/slang/Makefile.am.
      Fix a couple of picture repeat fields incorrectly compared to RepeatNormal.
      Fix fbCompositeTrans_0888xnx0888 build for wfb on big endian.
      EXA: Fix exaEnableDisableFBAccess for nested disables and enables.
      EXA: exaAssertNotDirty improvements.
      EXA: FillRegion{Solid,Tiled} improvements.
      EXA: PolyFillRect improvements.
      EXA: Glyphs improvements.
      EXA: Composite improvements.
      EXA: Defer to FillRegionTiled in Composite when possible.
      EXA: ImageGlyphBlt improvements.
      EXA: PutImage improvements.
      EXA: GetImage improvements.
      EXA: CopyNtoN improvements.
      EXA: SolidBoxClipped improvements.
      EXA: FillSpans improvements.
      EXA: Remove DrawableDirty.
      EXA: Fix OffscreenValidate build with DEBUG_OFFSCREEN enabled.
      DRI: Add TexOffset driver hooks.
      AIGLX: Zero-copy texture-from-pixmap.
      EXA: Export ExaOffscreenMarkUsed.
      Consolidate portPriv->pDraw assignments into xf86XVEnlistPortInWindow.
      mieq queue handling cleanups.
      dixLookupClient: Use access parameter.
      DRI: Fix build warning.
      DRITreeTraversal: Stop walking tree when we've seen all DRI windows.
      DRI: Clip cliprects obtained from DRIGetDrawableInfo to screen dimensions.
      Make sure BUILD_TIME doesn't have a leading zero.
      mieqEnqueue: Make local queue tail variables unsigned.
      exaGetImage: Don't migrate or try to accelerate for 1x1.
      exaPolyFillRect: Don't track damage explicitly.
      Make sure DRIScreenPrivIndex is -1 when no DRI screen private is allocated.
      Fix build when int10 doesn't use x86emu.
      Fix regression from recent composite changes.
      Make sure DRI drawables are cleaned up when client dies.
      GLX: Only build code dealing with GLXPixmap damage field when DRI is enabled.
      Fix RandR 1.2 conversion of two colour to ARGB cursor on MSB first platforms.

Nicolas Trangez (1):
      Hotplug: Remove unused function definition from hotplug.h

Peter Hutterer (22):
      dix/events: take screen number, not pointer, in PostSyntheticMotion
      fix: WarpCursor needs to send MotionNotify.
      mi:     Move WarpPointer event generation to miPointerMove to avoid duplicate
      Xext: Update device's lastx/lasty when sending a motion event with XTest.
      dix: remove 'register' keyword for all variables.
      mi: remove 'register' keywords.
      dix: Increase allocation size for core keyboard keymap to avoid buffer overrun when copying keymap from extension devices.
      Xi:     remove 'register' keywords.
      Documentation for events.c.
      Change dbus 'listDevices' call to not require an argument.
      config: Return errors as negative numbers, device ids as positive numbers.
      Use DevicePresence events to tell the client about enabled/disabled devices.
      Disable devices before removing, remove unrecoverable devices.
      Require inputproto 1.4.2.
      Check for identical grabs when adding a new passive grab. If an identical grab
      Split up memory for devices configured in the config file.
      Add a few comments to devices.c
      Only decrement buttonsDown when the button count is greater than 0.
      Set the detail field for DeviceKeyEvents to the keycode.
      NULL-terminate device list when synthesizing core devices.
      configFiles(): don't return anything when declared as void.
      Make sure window->optional is allocated before assigning it.

Remigiusz Marcinkiewicz (2):
      Input: Allow a pointer to a device to be returned in NIDR
      Config: Extend D-BUS API

Rich Coe (1):
      CheckConnections: don't close down the server client (bug #7876)

Samuel Thibault (5):
      kdrive: make building of Linux support conditional (bug #5613)
      mesa/indirect: include byteswap.h on GNU userland (bug #5613)
      xfree86/hurd: add bell support, remove SERVER/LOADER defines (bug #5613)
      xfree86/hurd: remove OsMouseProc (bug #5613)
      I/O enable/disable update for the Hurd

Soren Sandmann Pedersen (57):
      Fix for AMD cpu detection. Bug 9614, Dan Williams.
      Add new fbCompositeSrc_8888x0565mmx() function, based on patch by Dan
      Revert "Fix for AMD cpu detection. Bug 9614, Dan Williams."
      Reapply patch to fix AMD CPU detection
      Use the new 8888x0565mmx function in fbpict.c
      Merge David Reveman's gradient optimization patch from pixman
      Fix pixman bug 5777, patch by David Reveman
      Integrate David Turner's gradient optimizations from pixman
      Remove a few memory references in fbFetchTransformed
      Fixing gradient repeat mode computations in previous patch. From David
       Fix gradient walker to not reset needlessly
      Remove #if 0'ed leftovers from merge
      Gradient fixes
      Remove #if 0'ed merge leftovers
      Add functions fbCompositeSrcSrc_nxn() and fbCompositeTrans_0565xnx0565
      Fix format vs formatCode in previous commit
      From xserver via pixman (Jeff Muizelaar)
      From pixman (Jeff Muizelaar)
      Remove #if 0'ed blocks
      Use READ/WRITE macros for new functions introduced in previous commits.
      Various fixes from xserver via pixman (Billy Biggs)
      Add function fbCompositeSrcAdd_8888x8x8(), and fix a bug where
      Integrate optimization from xserver from David Reveman where repeats
      Port MSVC++ CPU detection code from pixman. (Vladimir Vukicevic).
      Don't treat convolution filters as transformations.
      More pixman merging
      Pixman merge
      Pixman merge
      Pixman merge
      Fix bug 8871 - scrolling corruption with a compositing manager
      Pixman merge
      Fix typo in previous commit
      Change expand_alpha_rev to expand_alpha in mmxSaturateU
      Bug fix in fbCompositeIn_nx8x8888
      Pixman merging
      Add fbCompositeRect() as another special case in the switch of doom in fbpict.c
      New fbWalkCompositeRegion() function
      Add dependency on pixman 0.9.0
      Turn boxes and regions into typedefs for pixman types
      Add new InitRegions() function called from dix/main
      Port large amounts of the region code to pixman
      Revert various fast path functions to their pre-pixman-merge state
      Use the pixman fixed point types and macros
      Use pixman types for transforms and vectors
      Use pixman short formats, revert the gradient_stop change
      Make fbFetch_b8g8r8() actually write the read value to the buffer
      Remove excessive unrolling in fbCompositeSrc_x888x8x8888() and fix bug where
      Make the general compositing code create a pixman image and call
      Break image_from_pict() into a few subfunctions.
      Move fbCompositeGeneral() to fbpict.c and remove fbcompose.c
      Comment out setup of general MMX code
      Use pixman_image_set_indexed() to make 8 bit work
      Remove fast path code from fbpict.c
      Add missing offsets for window coordinates - reported by Colin Harrison
      Port renderedge.c to pixman
      - Make image_from_pict() non-static
      Delete trapezoid rendering code; replace with pixman calls

Stefan Huehner (1):
      Bug #10560: Code-Cleanup: function declarations () -> (void)

Søren Sandmann Pedersen (12):
      Remove most of the fast-path MMX operations from fbmmx. fbCopyAreammx
      Delete fbCompositeCopyAreammx()
      Don't pass regions to pixman_image_composite() anymore.
      Split fbCopyAreammx() into a new fbBltmmx() function; call this
      Split fbSolidFillmmx() into a new FbFillmmx() function. Call that from fbFill().
      Remove fbCopyAreammx() and fbSolidFillmmx()
      Port a few forgotten fbSolidFillmmx()'es to fbFillmmx(). Use pixman_blt() instead of fbBltmmx()
      Delete fbBltmmx().
      Require pixman 0.9.2
      Replace fbFillmmx() with pixman_fill() and remove fbmmx.[ch]
      Remove fbmmx.[ch] files
      In fbFill() make sure the drawable is validated when pixman_fill() succeeds.

Thomas Hellstrom (3):
      Add an EXA driver callback to determine whether a pixmap is
      Changes for single-entity multi-screen DRI.
      libdri: Make sure the new DRIInfo keepFDOpen member is honoured.

Tiago Vignatti (9):
      Fix kdrive XKB.
      Janitor work. Remove 'defined but not used' warnings from xorg and other cosmetic.
      More janitor work. Remove 'defined but not used' warnings from kdrive and some minor cosmetic.
      Remove double-defined.
      Relink properly all kdrive servers when changes happen.
      For each kdrive server put a dependencie on its own libraries.
      Remove redundant linking in kdrive. Fix configure.ac variable name and clean
      kdrive must to know that devices are unplugged.
      Postpone options variable assignment to fix segfault when we got a device but

Tilman Sauerbeck (2):
      Distribute hw/xfree86/modes.
      Bug #10823: Fixed default OSNAME value.

Tomas Janousek (1):
      Bug #10296: Fix timer rescheduling.

Zephaniah E. Hull (6):
      Print the build time as well as the date if we can.
      Fully init the AbsoluteClassRec in InitAbsoluteClassDeviceStruct.
      Add xf86PostMotionEventP, takes a pointer instead of a variable number of
      Let's not do a calloc and a free on every call to xf86PostMotionEvents.
      xf86PostMotionEvents[P] calls xf86SendDragEvents, xf86SendDragEvents
      In NewInputDeviceRequest, only call EnableDevice if xf86Screens[0]->vtSema is

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://lists.freedesktop.org/archives/xorg-announce/attachments/20070801/3a21a78d/attachment-0001.pgp 

More information about the xorg-announce mailing list