If absolute events were posted, dixflags got set conditionally on whether the
valuators are different from the last posted set of values.
If dixflags are undefined however, the DIX interprets them as relative
valuators. Fix this by making sure defining dixflags is always defined.
X.Org Bug 17724 <http://bugs.freedesktop.org/show_bug.cgi?id=17734>
NIDR should be used to create a new SD from e.g. within a driver.
DIDR should be used to remove a device from the server.
Signed-off-by: Peter Hutterer <peter.hutterer@redhat.com>
Use ErrorF for an error message after an xcalloc failure, and return instead of falling through to GlxSetVisualConfigs, or abort()ing at the test branch.
(cherry picked from commit 1056700971)
If you need to bail out the server, use Ctrl-Alt-Fx, or enable zapping
if it bothers you that much. If Ctrl-Alt-Fx is broken, nag me until
it's permanently fixed.
The nvidia driver currently uses these hooks to work around problems where RAC
will disable access to the hardware at unexpected times. This change restores
these hooks until we can come up with a better API for working around RAC.
This reverts commit c1df4fbede.
The nvidia driver currently uses these callbacks to work around problems where
RAC will disable access to the hardware at unexpected times. This change
restores these hooks until we can come up with a better API for working around
RAC.
This reverts commit d7c0ba2e9e.
Conflicts:
hw/xfree86/loader/xf86sym.c
Some BIOSes (hi XGI!) will attempt to enumerate the PCI bus by asking
for the config space of every possible device number. This despite
perfectly functional BIOS methods to enumerate the bus exactly.
If anyone can come up with an example of a bus where:
- both i/o and memory resources are addressable
- access to them can be controlled
- but they can't be controlled independently
then by all means, reinstate this logic.
instead of calling CFRunLoopRun() directly. The leak wasn't reproducible on
this machine, but someone was able to produce a leak trace with Instruments
that indicates it was leaking in the CFRunLoopRun() path.
x-input.m: dequeue and ignore events when pbproxy_active is false.
x-selection.h: add an is_active method that is used by x-input.m to ignore
events.
x-selection.m: Handle nearly every preference, except for primary_on_grab,
which I don't really understand yet.
(cherry picked from commit 4d51ad851e)
Remove some unnecesssary headers.
Remove some dead code that was never called or used in pbproxy.
Make use of an NSAutoreleasePool in x_init. It could potentially cause a leak
on a startup without this.
Start adding reload_preferences to the x_selection class, as well as event
handling for that.
(cherry picked from commit 602e8ba8f7)
Fix the usage of the NSString cStringUsingEncoding: - it doesn't NUL
terminate the string, which lead to a bus error. So, we use
lengthOfBytesUsingEncoding: to get the length in bytes instead of
strlen().
(cherry picked from commit 6333d619e7)
2 of the paths leaked, when INCR transfers were done. Now we
are leak free according to the leaks program for all transfers
I have tried so far.
(cherry picked from commit aa98db576b)
NSAutoreleasePool. Now the usage is consistent. In x_input_run()
we create a pool, and release it after processing the XEvents.
Add some getpid() output to main for debugging. It needs a bit more
testing before the next release.
Don't retain the NSPasteboard as the old code did. That may have
contributed to the leak, and it made it so that we needed the
NSAutoreleasePool created in main().
Remove the _known_types, and _pasteboard instance variables from
the x_selection class. They aren't needed anymore.
The leaks program now indicates 0 leaks after some usage. I want
to test further, but this seems much better, and my memory usage
graph indicates it's not growing.
(cherry picked from commit b245d84a72)
macro causes a leak (according to the leaks program).
Attempt to fix several other leaks with release method calls.
For some reason the process still grows more than it should...
I will need to use some better methods than leaks, and malloc_history
I suspect. Whatever is leaking, it's hard to find. I need to isolate
the cases more.
Add a missing image/jpeg branch.
Remove read_prop_32 - it's not used.
(cherry picked from commit 63a680354d)
According to the press release:
Previous SGI contributions to the free and open source community
are now available under the new license. These contributions
include the SGI® OpenGL® Sample Implementation, the GLX™ API and
other GLX extensions.
[...]
"SGI has been one of the most ardent commercial supporters of free
and open source software, so it was important to us that we continue
to support the free software development community by releasing our
earlier OpenGL-related contributions under this new license," said
Steve Neuner, director of Linux, SGI. "This license ensures that all
existing user communities will benefit, and their work can proceed
unimpeded. Both Mesa and the X.org Project can continue to utilize
this code in free software distributions of GNU/Linux. Now more than
ever, software previously released by SGI under earlier GLX and SGI
Free Software License B is free."
"The GLX API" is here read to include the original GLX source release
from:
http://www.sgi.com/products/software/opensource/glx/download.html
ftp://ftp.sgi.com/sgi/opengl/glx/glx1_2.1.tgz
which includes glxext.c as included in XFree86, from which our copies
in glx/ and hw/dmx/glxProxy/ are derived.
Move the struct atom_list into the x-selection class, so that it's
no longer a global variable named atoms. This may ease pthread
integration and reduce the chances of symbols conflicting.
(cherry picked from commit c1403c713c)
We may need another branch to convert a PICT to a PNG or JPEG. For now
TIFF works well in all of the test image copying apps when converted to
PNG or JPEG with an NSBitmapImageRep class.
(cherry picked from commit adf339d8f9)
INCR transfers mixed with a CLIPBOARD change.
Fix a bug with some UTF-8 transfers. I can only guess that my UTF-8
tests before were inadequate. I can now copy/paste any characters between
uxterm and Textedit.app.
(cherry picked from commit 7ca1532e8e)
to verify that the behavior didn't change.
main.m: XInternAtom compound_text, and atom_pair.
pbproxy.h: Add compound_text and atom_pair to the struct atom_list.
x-selection.m: Add an #include of Xutil.h. Refactor the reply struct
initialization to be done in a common place. Add send_reply: to simplify
the code a bit more. Add send_compound_text: which handles the
COMPOUND_TEXT type. Add the beginnings of a send_multiple:. Change
handle_image:extension: to handle_image:. The extension: message isn't
needed anymore.
(cherry picked from commit 1e9460abdf)
We can now copy and paste images to and from X11. Text copying and
pasting works as well.
The NSPasteboard can contain TIFF or PICT images, and pbproxy will
translate to an image/png or image/jpeg request, and list those in the
TARGETS.
I added a description of the basic design at the top of x-selection.m.
I removed the request_data x_selection class struct. It's not needed.
(cherry picked from commit 4a8daf8846)
to a TIFFRepresentation for use with NSPasteboard.
This has been tested with the Gimp and works with some minor quartz-wm changes.
The Finder clipboard shows the image updates after an Edit -> Copy.
(cherry picked from commit 12912adaee)
This change adds some [self own_clipboard] calls in the necessary places to get the proper greedy behavior.
UTF8_STRING and STRING properties seem to work well now with the test cases (PRIMARY, and CLIPBOARD). I can copy from several different X apps, and have the behavior be correct when pasting. I also verified that quartz-wm isn't doing the copying, by disabling the quartz-wm paths.
(cherry picked from commit 934669f732)
Add TRACE() calls where appropriate to try to figure out why we are losing CLIPBOARD at times, after transferring PRIMARY to the NSPasteboard.
Use the new pbproxy_clipboard_to_pasteboard where appropriate.
(cherry picked from commit 40190675a6)
Convert the puts usage to use DB().
Add the initial handle_image method.
Check for nil in the NSString instantiation in various places.
Add some commentary to enhance the clarity of why I did some things.
(cherry picked from commit 37361567b6)
Under the terms of version 1.1, "once Covered Code has been published
under a particular version of the License, Recipient may, for the
duration of the License, continue to use it under the terms of that
version, or choose to use such Covered Code under the terms of any
subsequent version published by SGI."
FreeB 2.0 license refers to "dates of first publication". They are here
taken to be 1991-2000, as noted in the original license text:
** Original Code. The Original Code is: OpenGL Sample Implementation,
** Version 1.2.1, released January 26, 2000, developed by Silicon Graphics,
** Inc. The Original Code is Copyright (c) 1991-2000 Silicon Graphics, Inc.
** Copyright in any portions created by third parties is as indicated
** elsewhere herein. All Rights Reserved.
Official FreeB 2.0 text:
http://oss.sgi.com/projects/FreeB/SGIFreeSWLicB.2.0.pdf
As always, this code has not been tested for conformance with the OpenGL
specification. OpenGL conformance testing is available from
http://khronos.org/ and is required for use of the OpenGL logo in
product advertising and promotion.
The check can fail because the output from FBIOGET_VSCREENINFO is used to set
Clock in fbdev2xfree_timing(). Then in fbdevHWSetMode(), xfree2fbdev_timing()
is called which sets the pixclock based on Clock. The resulting circle results
in slight rounding errors, causing the comparision check in fbdev_modes_equal
to fail.
- Redo damage naming for more consistency.
- Call post submission functions only where appropriate.
- EXA can now live without it's odd damage workarounds.
stuck-modifier fixes (capslock)
3button-emulation now doesn't send the modifier key with the click
Added other options to fake_button2 and fake_button3 defaults options:
({l,r}{control,alt,command,shift})
(cherry picked from commit 8fb6a1cf44)
(cherry picked from commit ae9c1b3cfb)
No point warning about missing driver hooks, that just means the person
who gave you the driver is inept. Might as well just crash. Also,
just name anonymous screens as screen%d instead of failing after the 36th
screen. Bonus points if you can figure out what the failure mode would
be on the 36th screen, and what the effective screen limit was.
There is no way this code can have been building for anyone since pciaccess
was merged. BSD and Linux were already using OS code on sparc, the only
people who could want this are Solaris, who should be using pciaccess
anyway.
This code was effectively only used in ix86Pci.c to select PCI config
access type. Nobody should be using that path anymore, in the glorious
pciaccess world; kernel services should get it right for you.
the defaults from InitVelocityData() or hypothetic driver-side changes
are now respected, not overridden.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Readded the old exec() server startup path for regression testing.
Don't use the dynamic fd addition code since it's not quite working correctly.
(cherry picked from commit 08f3fe153e)
If the device doesn't have any BARs then it's just a stub for some
lame operating systems that need one PCI device per output for
multihead. No point in warning about it.
It's all a bit wonky since both sis(4) and xgi(4) claim to support the
Volari Z7 and V5/8 (0x0020 and 0x0040), so let's side with xgi(4), why
not. Note that the V3 (not V3XT) identifies itself as a trident chip.
Put out a warning if xorg.conf has InputDevice sections, but these aren't
referenced in the used ServerLayout. This is only performed if AllowEmptyInput
is enabled.
The reason behind this is that the server used to auto-add the first
mouse/keyboard sections if none where referenced. Now, with HAL and AEI
enabled by default, setups that relied on this auto-adding break and are left
without input devices. The least we can do is warn them.
Us shipping a GUI configuration utility (especially as part of the
server!) was pretty pointless. There was pretty much nothing it could
configure which wasn't already runtime adjustable: if you could get a
server up with functioning input and output, there wasn't much xorgcfg
could do for you.
Au revoir.
- Use a single common function to compute reducedness.
- Call it from both the old-school and new-school mode validation paths.
- Define monitor reduced-blanking support in accord with EDID 1.4.
- Attempt to filter RB DMT modes away from the "standard" EDID pool if
the monitor doesn't claim RB support.
On some panels you end up with all of:
- No range descriptor
- No description of physical connectivity
- Native panel size mode in standard timings list
In principle you're supposed to use the timings for that mode from the DMT
spec, but in practice the DMT spec has timings for both 1920x1200 normal
and 1920x1200RB, and the standard timing field gives you no way to
distinguish. And, of course, the non-RB timings don't fit in a single
DVI link.
A couple #if defined(Lynx) && defined(sun) had become just if defined(sun),
resulting in wrong settings for Solaris builds, so they're now just deleted.
OsInitColors always just returned TRUE, so just remove calls to it and
insane special-case logic. Remove unused kcolor.c implementation, and
merge oscolor.h into oscolor.c since it was the only user. Remove
open-coded strncasecmp in oscolor.c.
Since we no longer need to call OsInitColors after reading the config
file, just call PostConfigInit() from one place, and move PM handling to
one place so we can install the signal handlers earlier.
If devices are prepended to the list, their wake-up order on resume is not the
same as the original initialisation order. Hot-plugged devices, originally
inited last, are re-enabled before the xorg.conf devices and in some cases may
steal the device files. Result: we have different devices before and after
suspend/resume.
RedHat Bug 439386 <https://bugzilla.redhat.com/show_bug.cgi?id=439386>
- Allow returning multiple drivers to try for a given PCI id (for instance,
try "geode" then "amd" for AMD Geode hardware)
- On Solaris, use VIS_GETIDENTIFIER ioctl as well as PCI id to choose drivers
- Use wsfb instead of fbdev as a fallback on non-Linux SPARC platforms
Remove AEI check from configImpliedLayout as the setting isn't actually parsed
at this point anyway (written by Sasha Hlusiak).
Resurrect checkInput() and check for devices there if AEI is false (this also
creates the default devices if required).
Set AllowEmptyInput to enabled by default if hotplugging is enabled.
If no Screen is specified in the ServerLayout section, either take the first
one from the config file or autogenerate a default screen.
X.Org Bug 16301 <http://bugs.freedesktop.org/show_bug.cgi?id=16301>
<http://bugs.opensolaris.org/view_bug.do?bug_id=6685465>
This bug is caused by Xephyr not handling the RGB byte order correctly
of the server where Xephyr is displaying on. The previous code just
assumed that the order was RGB and did not take into account that
Xservers may use different order (such as BGR).
The fix is to add a function to calculate the byte order and bits
to shift based on the visual mask and the visual bits_per_rgb (which
is usually 8, but could be server dependent). Since the shifts won't
change once the display connection has been made, I can cache these
values so that Xephyr doesn't have to keep recalculating them everytime
it tries to translate the Xephyr colormap entries for Xephyr clients to
the actual server colormap entries (i.e. calling the function
hostx_set_cmap_entry() repeatedly for every colormap entry).
RandR 1.1 has a physical size for each mode. It used to be that the DIX would
remember these modes and pass them back up to the DDX when changing the screen
configuration. The DDX uses RR_GET_MODE_MM to query the driver for the physical
dimensions of the screen, allowing it to preserve the DPI.
With RandR 1.2, the physical dimensions are stored as part of the output, rather
than per mode. The DIX only uses the sizes passed in from the DDX to select the
mode pool for the "default" output, and forgets the physical sizes. Then, when
reconfiguring the screen, it makes up a new RRScreenSizeRec using the dimensions
from the output, screwing up the DPI.
This change works around this problem by ignoring the DIX and querying the real
size from the driver.
This reverts commit 76576c87b0.
which was an incorrect revert of previous ABI bumps. Those
responsible for the accidental ABI bumps in both directions
have all been sacked.
This allows xf86-input-mouse to build again, for example.
Spiritual revert of 1fa4de80fc. Intel's C
compiler claims to be gcc-compatible; if they're not defining the same
macros as gcc then that's their bug, not ours. Even if we were to do
this aliasing we should do it once and for all in servermd.h.
Use only %di to name the PCI register to read/write, rather than %edi.
DOS is only expecting the base PCI config space anyway, and the BIOS
might be using the high bits of %edi.
Yes, this is a 486+ instruction and thus not strictly legal in vm86
mode, but enough BIOSes use it (looking at you VIA) that we might as
well implement it.
This code hasn't been updated with anything even resembling what anyone is
shipping in nearly thirty months. It hasn't built out of the box since
7.1. Most of its features over AIGLX are accomplished with DRI2 and
friends.
In the single output enabled case we never enter the loop and test
never gets set and so we fail to match a good mode.
This was causing my 2560x1600 to end up at 2048x1536.
Modelled after the xfree86 code. Call miDCInitialize to init the SW rendering
engine, then take the pointers, store it in a xnest-local variable, and put
the xnest-specific sprite funcs in place. In the xnest sprite funcs, call
through to the mi sprite funcs after doing xnest-specific stuff.
The problem happens if Monitor/Card combo doesn't provide EDID info,
and the XFree86-VidModeExtension extension is used.
Signed-off-by: Peter Hutterer <peter@cs.unisa.edu.au>
With the MD/SD device hierarchy we need control over the generation of the
motion history as well as the conversion later before posting it to the
client. So let's not let the drivers change it.
No x.org driver currently uses it anyway, linuxwacom doesn't either so dumping
it seems safe enough.
Recording damage from other operations (e.g. creating a client damage record)
may confuse the migration code resulting in corruption.
Option "EXAOptimizeMigration" appears safe now, so enable it by default. Also
remove it from the manpage, as it should only be necessary on request in the
course of bug report diagnostics anymore.
GNU/kFreeBSD defines __FreeBSD_kernel__, but not __FreeBSD__.
Unify preprocessor conditionals between variable declaration and use.
Debian bug #482550.
During GetPointerEvents (and others), we need to access the last coordinates
posted for this device from the driver (not as posted to the client!). Lastx/y
is ok if we only have two axes, but with more complex devices we also need to
transition between all other axes.
ABI break, recompile your input drivers.
This copies over the files generated from mesa/src/mesa/glapi. There's
a corresponding mesa commit that makes it easy to generate the glapi files
straight into the xserver tree when the XML definitions change.
The only few files that are copied from mesa but aren't generated are
glapi.[ch] and glthread.[ch]. Everything in there is technically DRI
driver API and the whole setup is still a bit fragile, but it's not a new
problem.
The --with-mesa-source configure option is still around since other
parts of the server (XGL and DMX - grep for MESA_SOURCE) need that,
but for common case of building with GLX and AIGLX support, that
option is no longer needed.
Conflicts:
Xext/xprint.c (removed in master)
config/hal.c
dix/main.c
hw/kdrive/ati/ati_cursor.c (removed in master)
hw/kdrive/i810/i810_cursor.c (removed in master)
hw/xprint/ddxInit.c (removed in master)
xkb/ddxLoad.c
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.
Seeing as this code seems to be specific to OpenBSD I don't think
__x86_64__ should have been added there at all. It appears to have
been added wherever __amd64__ existed before which is wrong. I
think that part of the commit should be reverted but also all four of
the checks should be __OpenBSD__ && __amd64__ instead of two one
direction and two flipped.
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".
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.
Most of these drivers didn't work. ati was the only one that even came
close. The igs, ipaq, itsy, pcmcia, savage, sis530, trident, trio, ts300,
and vxworks directories have never built since modularisation, so clearly
no one can miss them.
It was removed and simplified some conditionals. We don't need test for
pDev->isMaster inside xf86CursorSetCursor() because only MD enters there.
In the last chunk, ScreenPriv fields were being assigned without need, so
that code was wrapped inside the conditional to avoid it.
I also tried to make the identation more sane in some parts that I touched.
Signed-off-by: Tiago Vignatti <vignatti@c3sl.ufpr.br>
Minor modification, part of the original patch led to cursors not being
updated properly when controlled through XTest.
Signed-off-by: Peter Hutterer <peter@cs.unisa.edu.au>
The only function that cat set SWCursor before xf86DeviceCursorInitialize()
is xf86InitCursor() when VCP and is created.
Signed-off-by: Tiago Vignatti <vignatti@c3sl.ufpr.br>
Signed-off-by: Peter Hutterer <peter@cs.unisa.edu.au>
Missing parameter caused event processing to go nuts when checking valuators.
X.Org Bug 15936 <http://bugs.freedesktop.org/show_bug.cgi?id=15936>
Signed-off-by: Peter Hutterer <peter@cs.unisa.edu.au>
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.
Use dummy config functions to replace those from config/config.c, and
therefore do not link Xprt with $CONFIG_LIB.
Works around an endlessly spinning loop in dix/dispatch.c::Dispatch()
(WaitForSomething() not waiting) when built with dbus, which was
causing Xprt to use 95% cpu.
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.
We need a manual call to SetCursor when we switch from SW to HW rendering and
the other way round. This way we display the new cursor after removing the old
one.
In addition, we only update the internal state for the VCP's sprite. This way,
when we switch back to HW rendering the state is up-to-date and wasn't
overwritten with the other sprite's state.
The second part is a hack. It would be better to keep a state for each sprite,
but then again we don't have hardware that can render multiple cursors so we
might as well do with the hack.
Switches back to HW cursors when sprites other than the VCP are removed.
The current state requires the cursor to change shape once before it updates
to SW / HW rendering (whatever is appropriate), e.g. by moving into a
different window. Until this is done, the cursor is invisible.
This patch only creates a Files section if required, so if no entries are
added, an empty Files section will not be created.
Signed-off-by: Peter Hutterer <peter@cs.unisa.edu.au>
LeaveVT/EnterVT cycles will free/realloc shadow frame buffers. Because of
this, the presense/absence of that data is insufficient to know whether
the screen function wrappers are necessary. Instead, the 'transform_in_use'
flag should be used.
This patch also adds 'xf86RotateFreeShadow' for drivers to use at LeaveVT
time to free the rotation data; it will be reallocated on EnterVT.
In DeleteInputDeviceRequest, leave the conf_idev (which is shared with
xf86ConfigLayout.input) alone for devices that were specified in the
ServerLayout section of the config file. This way, in the next server
generation we are left with what was the original config and can thus re-init
the devices.
This is an addon to 6d22a9615a, an attempt to
fix Bug 14418.
X.Org Bug 15645 <https://bugs.freedesktop.org/show_bug.cgi?id=15645>
X.Org Bug 14418 <https://bugs.freedesktop.org/show_bug.cgi?id=15645>
The previous check works in the master-branch, but doesn't work with MPX. We
actually copy the SD's information into the MDs public.devicePrivate, so we
need to explicitly check whether a device is a MD before freeing the module.
glcore gets linked with -ldl, -lpthread for s3tc and glapi
xserver needs
DLOPEN_LIBS - to dlopen the glcore dso
LD_EXPORT_SYMBOLS_FLAG - to export symbols for glcore to use
the ld flag is added to kdrive only when GLX is enabled, the net overhead for
Xephyr is ~155KB, could be reduced with --dynamic-list.
When starting up kdrive/fbdev, if the current framebuffer mode is sensible use
that unless told otherwise.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
XKB was disabled in 08928afb05, with the comment
"Disable XKB, as we can't yet use it". Seems like "yet" is over, running GNOME
and changing XKB settings seems to work in Xnest now.
X.Org Bug 10015 <https://bugs.freedesktop.org/show_bug.cgi?id=10015>
Use __libmansuffix__ instead of __oslibmansuffix__ which isn't getting
replaced, and rewrap some text to get __xservername__ replaced in the
description of Option "Accel" (cpp doesn't like the preceding quote).
This extension provided bug-compatibility with pre-X11R6, but has been
stubbed out in our server since 2006 to return BadRequest when you actually
asked for it.
(cherry picked from commit 4e2c6dbabdbbaaca213fd08edd422de15d0900cc)
required because of commit 7c0709a736,
which made requestingClient in dix specific to Xprint only.
Add to XPRINT_LIBS in hw/xprint/Makefile.am in front of
$(XSERVER_LIBS) to override definitions in libdix.la for standard xservers.
Follows 571206832d (providing -DXPRINT
to xprint subdirs).
Note it may be possible to restructure the code so that
requestingClient is stored elsewhere than in dix. See discussions
following http://lists.freedesktop.org/archives/xorg/2008-March/033844.html
If this is done it may be possible to revert this commit (if not 571206...).
-DXPRINT had only been set for Xprt in hw/xprint/Makefile.am
After commit 7c0709a736 it is also
required for ps/PsArea.c and PsFonts.c to ensure ‘requestingClient’ is
defined, so make it a global Xprint definition in configure.ac.
(cherry picked from commit 28a6719fd486d9a9cecad0b057d9ea7c59c66055)
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>
A few pieces of code were abusing this define for other purposes, which are
converted to #ifndef DEBUG instead. There should be no ABI consequences
to this change.
The rationale is that having the define in xorg-server.h also disables
assert() drivers, which is unexpected, and also difficult to avoid since
xorg-server.h is included in their config.h, and you can't put a #undef in
config.h. As for removing it from the server instead of moving it to an
internal header, we probably shouldn't have unnecessary assert()s in
critical server paths anyway, and if we do we could #define NDEBUG in the
specific cases needed.
Some pointer devices send key events [1], blindly getting the paired device
crashes the server. So let's check if the device is a pointer before we try to
get the paired device.
[1] The MS Wireless Optical Desktop 2000's multimedia keys are sent through
the pointer device, not through the keyboard device.
The jstk code for Joysticks is not used by any module, was never actually compiled and uses an API
that is deprecated these days.
No reason to keep it.
InitValuatorDeviceClass.
Add InitProximityClassDeviceStruct call to prepare for tablet support.
(cherry picked from commit 1bd980a5b114f5320360943214f8f9f23b29c1e3)
Get rid of glcontextmodes.[ch] from build, rename __GlcontextModes to
__GLXcontext. Drop all #includes of glcontextmodes.h and glcore.h.
Drop the DRI context modes extension.
Add protocol code to DRI2 module and load DRI2 extension by default.
Since there's no way to safely know how many blocks xf86DoEDID_DDC2 would
return, add a new xf86DoEEDID entrypoint to do that, and implement the
one in terms of the other.
The latter doesn't give you the option's value, it just tells you if
it's present in the configuration. So using Option "EXANoComposite" "false"
disabled composite acceleration.
This patch (and not setting HARDWARE_CURSOR_BIT_ORDER_MSBFIRST on big endian
platforms) fixes it for me with the radeon driver and doesn't break intel.
Correct patch this time :)
Should have done this in the first place. Since we're checking for the absence
of the get_crtc callback in the first place, we'll short circuit the later call
and disable the output, so the ugly "continue" block is unnecesary.
By adding a new output callback, ->get_crtc, xf86SetDesiredModes is able to
avoid turning off outputs & CRTCs if the current output<->CRTC mappings are the
same as the desired configuration. This helps avoid flickering displays at
startup time, which speeds things up a little and looks better.
Unless we check for vtSema before calling into the CRTC and output callbacks,
we may end up trying to access video memory that no longer exists, leading to a
crash. So if we don't have vtSema, return FALSE to the caller, indicating that
we didn't do anything.
Fixes#14444.
Actually more like in the mainline case, where the ideal mode happens to
be the very first aspect match on the first monitor. But let's not
split hairs.
The address written to 0xcf8 contains the PCI slot address to send the
config cycle to. However, we would ignore that and always send the
cycle to the device whose BIOS we were running. This breaks some
integrated graphics platforms that have explicit knowledge about the
system's host bridge, for example.
While the ScreenRec's notion of size in millimeters would get updates,
the RANDR 1.1 notion wouldn't, so your screen would appear to be square
and probably at some ludicrous DPI.
xserver and libpciaccess both need to open /dev/xf86, which can only
be opened once. I implemented pci_system_init_dev_mem() like Ian
suggested. This requires some minor changes to the BSD-specific
os-support code. Since pci_system_init_dev_mem() is a no-op on
FreeBSD this should be no problem.
