Merge branch 'xsync-transition-bug-59644' into for-keith

sync: split updating and triggering a counter up

Now that the brackets are always the nearest change points (regardless of
transition) we need to update the counters whenever we check for any updates.

Otherwise we end up with a situation where counter->value is out of date and
an alarm doesn't trigger because we're still using the value from last time
something actually triggered.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>

sync: always set the brackets (#59644)

The current code sets bracket_greater to the first trigger after the current
value, and bracket_less to the last trigger before the current value.

For example, the idle timer with three negative and three positive transitions
would set this:

         nt1       nt2                 nt3
|--------|------|--|------- idle --|---|--|-----> t
               pt1                pt2    pt3
bracket_less == nt2
bracket_greater == pt2

This is an optimization so we can skip code paths in the block/wakeup handlers
if the current value doesn't meet any of the trigger requirements. Those
handlers largely do a
   if (bracket_less is less than current value &&
       bracket_greater is greater than current value)
        return, nothing to do

However, unless the bracket values are updated at the correct time, the
following may happen:

                                      nt
|--------------|---------- idle ------|--------> t
               pt

In this case, neither bracket is set, we're past the pos transition and not
yet at the neg transition. idle may now go past nt, but the brackets are not
updated. If idle is then reset to 0, no alarm is triggered for nt. Likewise,
idle may now go past pt and no alarm is triggered.

Changing an alarm or triggering an alarm will re-calculate the brackets, so
this bug is somewhat random. If any other client triggers an alarm when the
brackets are wrongly NULL, the recalculation will set them this bug may not
appear.

This patch changes the behavior, so that the brackets are always the nearest
positive or negative transitions to the current counter value. In the example
above, nt will trigger a wakeup and a re-calculation of the brackets, so that
going past it in the negative direction will then cause the proper alarm
triggers.

Or, in Keith's words:

  Timer currently past a positive trigger
      No bracket values, because no trigger in range

  Timer moves backwards before the positive trigger
      Brackets not reset, even though there is now a trigger in range

  Timer moves forward past the positive trigger
      Trigger doesn't fire because brackets not set

Setting the LT bracket in this case will cause everything to get
re-evaluated when the sync value moves backwards before the trigger
value.

X.Org Bug 59644 <http://bugs.freedesktop.org/show_bug.cgi?id=59644>

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>

sync: if the idle time was reset, force alarms to trigger (#70476)

The time between the idle reset and the IdleTimeWakeupHandler to be called is
indeterminate. Clients with an PositiveTransition or NegativeTransition alarm
on a low threshold may miss an alarm.

Work around this by keeping a reset flag for each device. When the
WakeupHandler triggers and the reset flag is set, we force a re-calculation of
everything and pretend the current idle time is zero. Immediately after is the
next calculation with the real idle time.

Relatively reproducible test case: Set up a XSyncNegativeTransition alarm for
a threshold of 1 ms. May trigger, may not.

X.Org Bug 70476 <http://bugs.freedesktop.org/show_bug.cgi?id=70476>

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>

dix: provide accessor methods for the last device event time

And now that we have the accessors, localize it. No functional changes, just
preparing for a future change.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>

sync: supply the counter from IdleTimeBlockHandler

The main idletime counter has an initialized deviceid, might as well be
supplying it properly. Without this, we'd only ever check the XIAllDevices
counter, so the wait would never be adjusted for the device-specific triggers.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>

sync: always call BracketValues when recalculating upper/lower brackets

Both ServertimeBracketValues and IdleTimeBracketValues copy the value into
there SysCounter privates. Call it for a NULL set as well, so we don't end up
with stale pointers and we can remove the block/wakeup handlers.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>

sync: compress two if statements

No functional changes, just merges a > and == condition into a >= condition.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>

dix: only deliver for the current grab type

Use the grabtype to determine which type of event to send - all other events
are pointless and may result in erroneous events being delivered.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

dix: don't attempt to deliver an event for a different grabtype

For an active grab, grab->eventMask can be either the core or the XI1 mask.
With the overlap of event filters, calling DeliverOneGrabbedEvent(XI1) for a
ProximityOut event will trigger if the client has selected for enter events -
the filter is the same for both.

Thus, we end up delivering a proximity event to a client not expecting one.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

include: change grabtypes to start at 1

Avoid erroneous detection of an unset grabtype as CORE

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

dix: only allow button and key events to freeze a sync'd pointer

If a client calls XAllowEvents(SyncPointer) it expects events as normal until
the next button press or release event - that freezes the device. An e.g.
proximity event must thus not freeze the pointer.

As per the spec, only button and key events may do so, so narrow it to these
cases.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Xephyr: restore cursor visibility

  Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=69388

Commit c100211034ab69ce453a1644fb61c6808d7e3eda (dix: only show the cursor
if a window defines one (#58398)) broke the default cursor behaviour in
Xephyr (unless run with -retro). Restore the default cursor visibility
so that '-retro' or '-host-cursor' are not needed to have a visible
cursor.

Signed-off-by: Michele Baldessari <michele@acksyn.org>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
as of ba387cf21f7d95987211f75d8024601e7d64e322 "ephyr: Use host (HW) cursors
by default." this only applies if -sw-cursor is given on the cmdline.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Fix Xephyr compilation when DEBUG is enabled

When DEBUG is enabled Xephyr compilation fails:
ephyrdriext.c:343:133: error: 'is_ok' undeclared (first use in this
function)
     EPHYR_LOG("leave. is_ok:%d\n", is_ok);

Just reemove bogus is_ok variable.

Signed-off-by: Michele Baldessari <michele@acksyn.org>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

input: calloc minimization for xi2mask_new

There's no reason to do this as (nmasks + 2) callocs, and it's a
surprisingly hot path.  Turns out you hit this ~once per passive grab,
and you do a few bajillion passive grab changes every time you enter or
leave the overview in gnome-shell.  According to a callgrind of Xorg
with gnome-shell-perf-tool run against it:

Ir before: 721437275
Ir after:  454227086

Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Jasper St. Pierre <jstpierre@mecheye.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

hw/xwin: Fix for "glx: Remove screen number from __GLXconfig"

Fix compilation after commit c3c976f54c3c282d6fa6c8360688e036bc43d210 "glx:
Remove screen number from __GLXconfig"

Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Colin Harrison <colin.harrison@virgin.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>

Merge remote-tracking branch 'jeremyhu/master'

configure.ac: Include missing proto declarations in SDK_REQUIRED_MODULES

Commits a1d41e311c21e, 7d859bd87834d & 3ed2c6e11298c made extinit.h require
the XF86 Big Font, XRes & ScrnSaver proto headers, but failed to add them
to the SDK_REQUIRED_MODULES so pkg-config would find them for driver builds.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Keith Packard <keithp@keithp.com>

XQuartz: Dont't set screen on __GLXConfig

Fixes regression introduced by c3c976f54c3c282d6fa6c8360688e036bc43d210

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>

rootless: Use miCopyRegion instead of fbCopyRegion

Fixes regression introduced by e657635dbe6b92875b0e88370557c2cbab673a49

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>

os: move <arpa/inet.h> for any !win32 system

It is needed in IPv6 configurations (for inet_pton) also when
SIOCGIFCONF is not defined.

Signed-off-by: Pino Toscano <toscano.pino@tiscali.it>
Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Signed-off-by: Julien Cristau <jcristau@debian.org>

xfree86/hurd: include <hurd.h>

Needed for using get_privileged_port.

Signed-off-by: Pino Toscano <toscano.pino@tiscali.it>
Acked-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Signed-off-by: Julien Cristau <jcristau@debian.org>

Bump version to 1.14.99.2

Signed-off-by: Keith Packard <keithp@keithp.com>

kdrive/ephyr: Remove duplicate ephyrExtensions and ephyrExtensionInit decls

These were duplicated when GLX support was re-added on two different branches.

Signed-off-by: Keith Packard <keithp@keithp.com>

xfree86: add a comment as to why the logdir is created

Without the logdir, the xserver will write the content of the log file on the
terminal stating that it cannot be written and will stop.

Refer to https://bugs.freedesktop.org/show_bug.cgi?id=3889

Reviewed-By: Matt Dew <marcoz@osource.org>
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Keith Packard <keithp@keithp.com>

xfree86: Use $(MKDIR_P) for better code portability

Still true that we should not use the lower case $(mkdir_p) version.
However, remove the 2005 comment as the MKDIR_P is widely used now.

Reviewed-By: Matt Dew <marcoz@osource.org>
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Keith Packard <keithp@keithp.com>

Uninstall X link and CYGWIN libXorg.exe.a in local install targets

It is our duty to uninstall any files and/or directories that we installed
through install-data-local and install-exec-hook.

Currently the X symbolic link to Xorg remains on disk after running
make uninstall.

Note the exception for logdir which is usually shared by other modules.

Reviewed-By: Matt Dew <marcoz@osource.org>
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Keith Packard <keithp@keithp.com>

Xorg binary: use install-exec-hook rather than install-exec-local

The former was explicitly designed to execute additional code after the binary
has been installed. The latter can be executed in any order, hence it's
current dependency on install-binPROGRAMS as a workaround.

The CYGWIN libXorg.exe.a target is an installation target rather than
a post-installation one, so it should not be done as a hook. It does not depend
on the Xorg executable being installed.

Automake:
"These hooks are run after all other install rules of the appropriate type,
exec or data, have completed. So, for instance, it is possible to perform
post-installation modifications using an install hook".

"With the -local targets, there is no particular guarantee of execution order;
typically, they are run early, but with parallel make, there is no way
to be sure of that".

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Keith Packard <keithp@keithp.com>

Use $(LN_S) provided by AC_PROG_LN_S macro to create links

For better code portability.

Reviewed-By: Matt Dew <marcoz@osource.org>
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Keith Packard <keithp@keithp.com>

The Xorg binary is missing the extension $(EXEEXT) in the makefile

This is not a problem on UNIX platforms, but on CYGWIN it creates a broken
link to Xorg rather than a link to Xorg.exe.

From the CYGWIN log on tinderbox, we can see that the executable Xorg.exe is
installed correctly. We can see the command used to create the link:

(cd /jhbuild/install/[...]/install/bin && rm -f X && ln -s Xorg X)

Note that the "relink" makefile target correctly appends $(EXEEXT) to Xorg.

Reviewed-By: Matt Dew <marcoz@osource.org>
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Keith Packard <keithp@keithp.com>

Merge remote-tracking branch 'anholt/ephyr-fixes'

miext/shadow: missing c2p_core.h breaks "make distcheck" target.

Reviewed-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Keith Packard <keithp@keithp.com>

Merge remote-tracking branch 'whot/for-keith'

Merge remote-tracking branch 'ajax/xserver-next'

test: add new os executable to .gitignore

Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

xserver: enable InputClass option "GrabDevice" by default for non-seat0 seats (#69478)

This patch contributes to fill the remaining gaps which make
systemd-multi-seat-x wrapper still necessary in some multiseat setups.

This also replaces previous evdev patch that does the same thing
for that particular driver.

When option "-seat" is passed with an argument different from "seat0",
option "GrabDevice" for input devices is enabled by default
(no need of enabling it in xorg.conf's "InputClass" section).

Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=69478

Signed-off-by: Laércio de Sousa <lbsousajr@gmail.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

dmx: provide enough space for axis mappings

relmap/absmap is used as a evdev-axis-to-x-axis mapping. ABS_X maps to
axis 0, ABS_Y to 1, etc. skipping over non-existing axes so that the third bit
set in the ABS_* range is axis 2, and so on. This requires us to actually have
enough space to have all the ABS_*/REL_* range.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>

damage: Must translate initial window damage by window offset

Damage is reported relative to the drawable origin, but the window
borderClip is absolute. Translate the region by the window position
before reporting damage to adjust.

Reported-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>

XQuartz: pbproxy: Fix build with -DDEBUG

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>

XQuartz: Fix build with moved pseudoramiX

Regression from: e716baedc4d8e52a60f43ef21aba771b340d8c8b

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
Reviewed-by: Jon TURNEY <jon.turney@dronecode.org.uk>

glx: Fill in some missing attributes from DoGetFBConfigs

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Remove a dead comment

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Catch another failure case in drawable creation

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Remove screen number from __GLXconfig

Not used.  There's no real reason to match against this instead of
matching against fbconfig or visual ID anyway.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Remove pixmapMode from __GLXconfig

This has never been filled in with anything meaningful afaict, and you
can't get to it from the client in any event.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Remove support for NV_vertex_program and NV_fragment_program

Mesa doesn't implement these anymore, never really did outside of swrast
anyway.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: realloc style fix in RenderLarge

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Eliminate a small malloc from QueryContext

No reason to have that be a failure path.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Handle failure to create the pixmap backing the pbuffer

We happen not to sanitize the width/height we pass to CreatePixmap here,
oops.  It's not exploitable, but it's certainly a crash, so let's just
throw BadAlloc instead.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Implement GLX_PRESERVED_CONTENTS drawable attribute

We back pixmaps with pbuffers so they're never actually clobbered.  Say
so when asked.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Implement GLX_FBCONFIG_ID in GetDrawableAttributes

Required by GLX 1.4, section 3.3.6, "Querying Attributes".

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Implement GLX_{WIDTH,HEIGHT} in GetDrawableAttributes

Required by GLX 1.4, section 3.3.6, "Querying Attributes".

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Compute number of attributes in GetDrawableAttributes on the fly

This doesn't have any effect yet, but is needed to properly build the
reply for pbuffers.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Style fixes

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Pull GLX vendor string out of __GLXscreen

Given how we're currently implementing GLX this can't meaningfully vary
per-screen.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glx: Remove unused bits from the context struct

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

glxproxy: Don't track GLCLientm{aj,in}orVersion

Basically just a port of 62f06b0d to glxproxy.

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

dmx/glx: Remove unused __glXNop

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

damageext: Style fix

Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

damage: Remove "post-rendering" hooks

This is a revert of 974db58f5b730c3770ee461665a02dd4334d1dea.  There are
no consumers of this API, we must not need it.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

damage: Don't rewrite Text ops to GlyphBlt ops

There's no particularly good reason to, and it breaks Xnest.

Bugzilla: http://bugs.freedesktop.org/2454
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

damage: Simplify DamageUnregister

You can only register one drawable on a given damage, so there's no
reason to require the caller to specify the drawable, the damage is
enough.  The implementation would do something fairly horrible if you
_did_ pass mismatched drawable and damage, so let's avoid the problem
entirely.

v2: Simplify xf86RotateDestroy even more [anholt]

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

damage: Implicitly unregister on destroy

There's no reason not to, and it simplifies quite a few callers.

Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

fb: Remove unused compatibility wrappers

Originally added in early 2009, not being used elsewhere anymore.

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

mipointer: Remove deprecated miPointerCurrentScreen

The only remaining use was in some debugging code in DMX.

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Mark Kettenis <kettenis@openbsd.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>

xfree86: Remove deprecated unimplemented xf86MapReadSideEffects

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Mark Kettenis <kettenis@openbsd.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>

dix: FIXES is not optional

It's already not optional at configure time, this just makes it so at
build time too.

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

vbe: Don't try to load the ddc submodule

DDC is built into the server now.

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

misprite: Delete some dead private storage

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

dri2: Don't bother with xf86LoaderCheckSymbol("DRI2Connect")

The DRI2 code is now built-in to the server, even for Xorg.  The only
thing this could protect against is trying to run a libglx built with
DRI2 support against an Xorg built without it, which is firmly in
"doctor it hurts when I do this" territory.

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>

list: Avoid using X types

In particular, Bool.  This is not an ABI break:

/usr/include/X11/Xdefs.h:typedef int Bool;

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

dbe: Fold the window private private into the window private

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

dbe: Fold midbe reset into dix

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

dbe: Drop the unused idiom support

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

damage: Remove the 'damage window' resource type.

Note that the existing code never actually creates a resource of this
type, so it can't possibly do anything.  This is clearly meant to track
window destruction and clean up any associated Damages, but that's
already handled by miext/damage's DestroyWindow wrapper.

Previous discussion:
http://lists.freedesktop.org/archives/xorg-devel/2011-March/020847.html

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

dmx: Remove ShadowFB support

This has been listed as deprecated ever since DMX was merged.

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

dix: Remove a dead parameter from DoGetImage

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

mipointer: Remove EnqueueEvent from miPointerScreenFuncRec

No DDX overrode this, and we never actually called through that slot
anyway.

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

mipointer: Flatten calls to mieqSwitchScreen

No DDX was overriding this.

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

os: Factor out some common code in input buffer handling

Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>

xnest: Ignore GetImage() error in xnestGetImage()

When an Xnest instance is not viewable it will crash when a client in
that instance calls GetImage. This is because the Xnest server will
itself receives a BadMatch error.
This patch ignores the error. The application which has requested the
image will receive garbage - this however is fully legal according
to the specs as obscured areas will always contain garbage if there
isn't some sort of backing store as discussed in
https://bugs.freedesktop.org/show_bug.cgi?id=9488
The applied patch is a version from Dadek Doulik.

v2: Call XSync() before changing error handlers as suggested by
    Daniel Stone <daniel@fooishbar.org>.

Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Egbert Eich <eich@freedesktop.org>

Xnest: Implement xnestModifyPixmapHeader

Xnest variant of ModifyPixmapHeader that creates new Pixmap in parent X
server if it's size is modified from 0x0 to anything bigger.

xnestCreatePixmap doesn't create pixmap in parent X server if it has
dimensions 0x0. If it is later resized and accessed, Xnest will be
aborted with BadDrawable error from parent X server because it will
use XID 0. This happens with ScratchPixmap, for example as used from
XaceCensorImage. Applications using XACE crash Xnest.

Reviewed-by: Adam Jackson <ajax@redhat.com>

security: Don't count RetainPermanent clients twice

If a RetainPermanent client is subsequently killed by a KillClient
request, the reference count is decremented twice. This can cause the
server to prematurely kill other clients using the same Authorization.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Peter Harris <pharris@opentext.com>

mi: Avoid stack smash when drawing dashed lines

X.org Bug 54013 <https://bugs.freedesktop.org/show_bug.cgi?id=54013>

Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Peter Harris <pharris@opentext.com>

os: Immediately queue initial WriteToClient

If we immediately put the WriteToClient() buffer into the socket's write
queue, not only do we benefit from sending the response back to client
earlier, but we also avoid the overhead of copying the data into our own
staging buffer and causing extra work in the next select(). The write is
effectively free as typically we may only send one reply per client per
select() call, so the cost of the FlushClient() is the same.

shmget10:   26400 -> 110000
getimage10: 25000 -> 108000

shmget500:   3160 -> 13500
getimage500: 1000 -> 1010

The knock-on effect is that on a mostly idle composited desktop, the CPU
overhead is dominated by the memmove in WriteToClient, which is in turn
eliminated by this patch.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

ephyr: restore GLX support

It was removed since version 1.13

Signed-off-by: Sebastien Bacher <seb128@ubuntu.com>
Acked-by: Daniel Stone <daniel@fooishbar.org>

miext/sync: Fix multi-screen support in SyncFence initialization

miSyncSetup was checking to see if the screen private key had been
registered, and if so, skipping the setup of the provided screen.

Instead, it should register the private index only once, but then
initialize the screen unless it has already been initialized. This
latter step allows drivers to initialize the sync private structures
before the Sync extension itself is initialized.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>

randr: Fix a copypasta bug in CRTC confinement

Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>

Merge remote-tracking branch 'jturney/master'

Fix typo in configure warning.

Signed-off-by: Thomas Klausner <wiz@NetBSD.org>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

Fix bug in cursor handling.

CreateCursor (Xlib call XCreatePixmapCursor) with a non-bitmap
source pixmap and a None mask is supposed to error out with BadMatch,
but didn't.

From der Mouse <mouse@Rodents-Montreal.ORG>, changed following
comments by Alan Coopersmith <alan.coopersmith@oracle.com>.

Signed-off-by: Thomas Klausner <wiz@NetBSD.org>
Reviewed-by: Jasper St. Pierre <jstpierre@mecheye.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

kdrive: fix build error on gcc 4.8 for out-of-bounds array access

I'm getting a error building xorg-server-1.14.1.902 with thelatest snapshot
of gcc-4.8:

input.c:225:43: error: array subscript is above array bounds
[-Werror=array-bounds]

This is because kdNumInputFds can become equal to KD_MAX_INPUT_FDS in
KdRegisterFd(). This means that in KdUnregisterFd(), kdInputFds[j + 1] can
be beyond the end of the array.

Signed-off-by: Chris Clayton <chris2553@googlemail.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>

kdrive: initialize GLX for xephyr

Like commit ac1a60e7b6f06fd075cc5bf55d6bc67206a01d29, re-add
initialization of GLX after it was accidentally dropped from non-Xorg
servers in 5f5bbbe543f65c48ecbb5cce80116a86ca3fbe86.

Signed-off-by: Sebastien Bacher <seb128@ubuntu.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=62346
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Eric Anholt <eric@anholt.net>

ephyr: Do grab/ungrab for ctrl+shift, not just shift+ctrl.

Given that the window title says "ctrl+shift", having pressing those
keys in that order not ungrab you is fairly mean.

Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>

ephyr: Flush the X connection when updating the window title.

Otherwise when you're doing the ctrl-shift mouse grab thing, you
don't know what state you're in until the next rendering occurs.

Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>

ephyr: Use host (HW) cursors by default.

Unless you're working on the sw cursor rendering code, you surely want
to have real hardware cursors.

Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>

ephyr: Garbage collect some DOA host window clipping code.

Introduced in 79782726617d4b5a4f9b376f21936d035fc870e1 but never used.

Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Julien Cristau <jcristau@debian.org>

ephyr: Garbage collect some dead XV clipping code.

Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Julien Cristau <jcristau@debian.org>

ephyr: Remove the helper libs for each of the optional components.

Note that EXTRA_DIST was always unnecessary, because automake notices
conditional compile of source files and includes them.

Copyright header is added because git noted that this was a 61%
rewrite.

Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Julien Cristau <jcristau@debian.org>

ephyr: Remove the remaining bits of host/server XV split.

Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Julien Cristau <jcristau@debian.org>