* Make sure available areas are considered to have no eviction cost. This seems
to help for https://bugs.freedesktop.org/show_bug.cgi?id=15513 but I'm afraid
that may just be coincidence.
* Only calculate eviction cost of each area once for each eviction pass.
Safeguard against potential (though unlikely) division by zero.
* Cosmetic enhancements: Name eviction cost related variables 'cost' instead of
'score' to emphasize that smaller values are better, update Doxygen file
comment to the way eviction works now.
(cherry picked from commit 6c95fae1e9)
This allows using PanoramiXTranslateVisualID to test whether a given visual made
it through PanoramiXVisualConsolidate.
(cherry picked from commit 8b3d26f5b6)
If the monitor isn't reduced-blanking (either through EDID logic, or
config file setting), then remove RB modes from the default pool. Any
RB modes from the driver and config file pools will stick around though;
you asked for them, you got them.
(cherry picked from commit 0178b6a4ab)
The -wm (when mapped) option for the BackingStore support has been
causing the server to dereference a NULL pointer.
This has probably been the case since backing store has been
implemented on top of Composite.
It looks like (some of?) Composite didn’t expect its WIndowPtr
argument to be the root window.
In Composite’s compCheckRedirect() function we now avoid calling
compAllocPixmap() and compFreePixmap() when the pWin pointer’s
parent member is NULL, as is it the case with a server’s root window.
This addresses:
https://bugs.freedesktop.org/show_bug.cgi?id=15878
(cherry picked from commit 04211c3532)
The first guess used to be "is the preferred mode for one output the
preferred mode on all outputs". Instead, do "find the largest mode that's
preferred for at least one output and available on all outputs".
(cherry picked from commit 459f34b089)
Old logic was just the first one that happened to have an associated
CRTC. The new logic tries to find one that's definitely connected, has
probed modes, and has the largest candidate mode.
(cherry picked from commit 96111c1547)
The first fbconfig which has a depthbuffer > 0 and doublebuf is choosen
when associating fbconfigs with the visuals, indepenent of stencil bits.
This happens to work ok on intel as there all fbconfigs with a
depthbuffer > 0 also have stencil bits.
This patch fixes this by first trying to get a fbconfig for default X visuals
with both stencilbuf, depthbuf and doublebuffering, and if that fails fallback
to trying to get one with only a depthbuf and doublebuffering.
(cherry picked from commit f6e22d69af)
Create a new exported global variable, XineramaVisualsEqualPtr. Use this
pointer to decide whether two visuals are equal during visual consolidation.
This pointer can be wrapped, which allows drivers and extensions to control
which visuals are consolidated. A wrapper can reject the visuals without
calling down, but must call down and return that result if it deems the visuals
equal. This ensures that all layers agree that the visuals are equal.
Pass the screen of the other visual into the VisualsEqual callchain.
Don't free PanoramiXVisuals since we need it for PanoramiXTranslateVisualID.
Don't skip the first visual on the other screen in PanoramiXMaybeAddVisual.
Skip the loop in PanoramiXTranslateVisualID if screen is 0.
(cherry picked from commit c50b5d9789)
KdInitOutput() used to enable Composite when it was disabled by default,
but now this hack prevents ``-extension Composite'' from working.
Remove it, as Composite is enabled by default anyway.
(cherry picked from commit 9dfb525f6c)
Since glyphs are stored in pixmaps now, they can make their way into VRAM,
which invalidates a bunch of fast-path assumptions in the XAA code. Thus
you end up doing color-expands or WriteBitmap from la-la land and your
aliased glyphs go all funny.
Since XAA isn't ever growing the ability to do sane glyph accel, just force
glyph pixmaps into host memory by catching them at CreatePixmap time.
(cherry picked from commit 718652eaf9)
When something went wrong building a keymap, try to explain to the user
what it actually was, instead of the dreaded 'Failed to load XKB keymap'
catch-all.
(cherry picked from commit cf20df39cc)
Thanks to Owen Taylor for root-causing this one.
If a TreatAsTransparent window has any area in the borderClip, that will be
added to the totalClip region for use by other windows. That's wrong.
Instead, simply empty the borderClip for TreatAsTransparent windows right up
front.
(cherry picked from commit 6c1accce87)
I don't think this is the 100% correct answer as I get log spam saying
(EE) DoSwapInterval: cx = 0x98b8998, GLX screen = 0x96dd780
(EE) AIGLX: cx->pGlxScreen->swapInterval == NULL
but thats better than X exiting in my book.
(cherry picked from commit a368ab757e)
I'm not sure this the complete proper solution, perhaps it should explicitly fill
in ever field.
This at least makes glxinfo on glcore return sensible information, it doesn't make
gears work yet though.
(cherry picked from commit c61087c827)
The DDX (xfree86 anyway) maintains its own device list in addition to the one
in the DIX. CloseDevice will only remove it from the DIX, not the DDX. If the
server then restarts (last client disconnects), the DDX devices are still
there, will be re-initialised, then the hal devices come in and are added too.
This repeats until we run out of device ids.
This also requires us to strdup() the default pointer/keyboard in
checkCoreInputDevices.
X.Org Bug 14418 <http://bugs.freedesktop.org/show_bug.cgi?id=14418>
(cherry picked from commit 6d22a9615a)
The top bit of 0xCF8 is an enable bit, not part of the domain. Sending
cycles to domain 128 instead of domain 0 is rarely the right thing to do.
(cherry picked from commit cc7c045bae)
For non-Linux, _POSIX_C_SOURCE and friends restrict symbols defined rather
than enabling defines of symbols. Additionally, CLOCK_MONOTONIC was
apparently added to the standard around 2000 anyway, not 1993.
(cherry picked from commit d1de3dda8e)
Reduce the cost of the inner loop, by keeping a set of pointers to the
first and the last areas in the series, subtracting the cost of the first
area from the score, and adding the cost of the last area while walking
the list. This commit also moves the scanning loop from exaOffscreenAlloc
into a separate function.
Idea by Michel Dänzer.
(cherry picked from commit 8074676d2d)
Replace the current score keeping algorithm with a rolling counter that's
incremented in ExaOffscreenMarkUsed, with the previous value being stored
in the area. exaOffscreenAlloc uses the difference between the counter
value and the value in the area when deciding which area to evict.
It now also takes the size of the areas into account, and favors evicting
smaller areas.
The credit for these ideas goes to Michel Dänzer.
(cherry picked from commit 93d876891d)
