For predefined resource types, the offset of the devPrivates field was
already kept in a constant table. The only non-predefined type needing
this treatment was dbeDrawableResType, which is just a magic alias for
RT_PIXMAP.
This patch special-cases looking up RC_DRAWABLE offsets and uses the
table directly for everything else.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jamey Sharp <jamey@minilop.net>
This makes all of the previous macros into inline functions and also
turns all of the direct calls to pixman region code into inline
functions as well.
Reviewed-by: Jamey Sharp <jamey@minilop.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
This is a combination of a huge mechanical patch and a few small
fixups required to finish the job. They were reviewed separately, but
because the server does not build without both pieces, I've merged
them together at this time.
The mechanical changes were performed by running the included
'fix-region' script over the whole tree:
$ git ls-files | grep -v '^fix-' | xargs ./fix-region
And then, the white space errors in the resulting patch were fixed
using the provided fix-patch-whitespace script.
$ sh ./fix-patch-whitespace
Thanks to Jamey Sharp for the mighty fine sed-generating sed script.
The hand-done changes involve removing functions from dix/region.c
that duplicate inline functions in include/regionstr.h, along with
their declarations in regionstr.h, mi.h and mispans.h.
Reviewed-by: Jamey Sharp <jamey@minilop.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
This prepares the file to be moved from mi to dix. This patch
was done mechanically with the included scripts 'fix-miregion' run over
the entire X server and 'fix-miregion-private' run over
include/regionstr.h and mi/miregion.c.
Reviewed-by: Jamey Sharp <jamey@minilop.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
Since reallocating the backing pixmap can fail, we need to try and do
it before any other side effects of reconfiguring the window happen.
This changes the ConfigNotify hook to return status, and moves the
composite window reconfiguration wrappers to ConfigNotify. They all
basically did the same thing, so we can drop the MoveWindow,
ResizeWindow, ChangeBorderWidth wrappers, and allow ConfigNotify to do
all the work. If reallocation fails we fail before we send any
confiureNotify events, or enter the area we can't recover from.
The only place we now enforce 32k limits are in EXA/UXA/fb, so drivers
that don't use this should probably deal with it in their pixmap
allocate if they don't already.
This also breaks ABI, so we need an alternate fix for older servers,
working on the X server makes me realise why I'm a kernel hacker.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
This hides a MAXSCREENS-sized array as an implementation detail of
panoramiX.c rather than an exported global.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Tiago Vignatti <tiago.vignatti@nokia.com>
Tested-by: Tiago Vignatti <tiago.vignatti@nokia.com> (i686 GNU/Linux)
Many references to the dixScreenOrigins array already had the
corresponding screen pointer handy, which meant they usually looked like
"dixScreenOrigins[pScreen->myNum]". Adding a field to ScreenRec instead
of keeping this information in a parallel array simplifies those
expressions, and eliminates a MAXSCREENS-sized array.
Since dix declared the dixScreenOrigins array, I figure allocating a
screen private for these values is overkill.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Tiago Vignatti <tiago.vignatti@nokia.com>
Tested-by: Tiago Vignatti <tiago.vignatti@nokia.com> (i686 GNU/Linux)
Many references to the WindowTable array already had the corresponding
screen pointer handy, which meant they usually looked like
"WindowTable[pScreen->myNum]". Adding a field to ScreenRec instead of
keeping this information in a parallel array simplifies those
expressions, and eliminates a MAXSCREENS-sized array.
Since dix uses this data, a screen private entry isn't appropriate.
xf86-video-dummy currently uses WindowTable, so it needs to be updated
to reflect this change.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Tiago Vignatti <tiago.vignatti@nokia.com>
Tested-by: Tiago Vignatti <tiago.vignatti@nokia.com> (i686 GNU/Linux)
Most references to the savedScreenInfo array already had the
corresponding screen pointer handy, which meant they usually looked like
"savedScreenInfo[pScreen->myNum]". Adding a field to ScreenRec instead
of keeping this information in a parallel array simplifies those
expressions, and eliminates a MAXSCREENS-sized array.
Since dix uses this data, a screen private entry isn't appropriate.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Tiago Vignatti <tiago.vignatti@nokia.com>
Tested-by: Tiago Vignatti <tiago.vignatti@nokia.com> (i686 GNU/Linux)
Makes the use of IsMaster in ProcChangeKeyboardControl consistent with other
similar loops.
Signed-off-by: Nicolas George <nicolas.george@normalesup.org>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
For absolute input devices (E.G. touchscreens) in multi-head setups,
we need a way to bind the device to an randr output. This adds the
infrastructure to the server to allow us to do so.
positionSprite() scales input coordinates to the dimensions of the shared
(total) screen frame buffer, so to restrict motion to an output we need to
scale/rotate/translate device coordinates to a subset of the frame buffer
before passing them on to positionSprite.
This is done here using a 3x3 transformation matrix, which is applied to
the device coordinates using homogeneous coordinates, E.G.:
[ c0 c1 c2 ] [ x ]
[ c3 c4 c5 ] * [ y ]
[ c6 c7 c8 ] [ 1 ]
Notice: As input devices have varying input ranges, the coordinates are
first scaled to the [0..1] range for generality, and afterwards scaled
back up.
E.G. for a dual head setup (using same resolution) next to each other, you
would want to scale the X coordinates of the touchscreen connected to the
both heads by 50%, and translate (offset) the coordinates of the rightmost
head by 50%, or in matrix form:
left: right:
[ 0.5 0 0 ] [ 0.5 0 0.5 ]
[ 0 1 0 ] [ 0 1 0 ]
[ 0 0 1 ] [ 0 0 0 ]
Which can be done using xinput:
xinput set-prop <left> --type=float "Coordinate Transformation Matrix" \
0.5 0 0 0 1 0 0 0 1
xinput set-prop <right> --type=float "Coordinate Transformation Matrix" \
0.5 0 0.5 0 1 0 0 0 1
Likewise more complication setups involving more heads, rotation or
different resolution can be handled.
Signed-off-by: Peter Korsgaard <peter.korsgaard@barco.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Peter wants to get a larger patch sequence put together and I didn't
read past the commit message to see the 'don't take this patch
please'.
This reverts commit 531ff40301.
Some input drivers need to implement an internal hotplugging scheme for
dependent devices to provide multiple X devices off one kernel device file.
Such dependent devices can be added with NewInputDeviceRequest() but they are
not removed when the config backend calls DeleteInputDeviceRequest(),
leaving the original device to clean up.
Example of the wacom driver:
config/udev calls NewInputDeviceRequest("stylus")
wacom PreInit calls
NewInputDeviceRequest("eraser")
NewInputDeviceRequest("pad")
NewInputDeviceRequest("cursor")
PreInit finishes.
When the device is removed, the config backend only calls
DeleteInputDeviceRequest for "stylus". The driver needs to call
DeleteInputDeviceRequest for the dependent devices eraser, pad and cursor to
clean up properly.
However, when the server terminates, DeleteInputDeviceRequest is called for
all devices - the driver must not remove the dependent devices to avoid
double-frees. There is no method for the driver to detect why a device is
being removed, leading to elaborate guesswork and some amount of wishful
thinking.
Though the input driver's UnInit already supports flags, they are unused.
This patch uses the flags to supply information where the
DeleteInputDeviceRequest request originates from, allowing a driver to
selectively call DeleteInputDeviceRequest when necessary.
Also bumps XINPUT ABI.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Dan Nicholson <dbn.lists@gmail.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
No special memory handling is used to give drivers the maximum flexibility
with the data. Drivers should be able to call realloc on the product string
if needed and perform similar operations.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Dan Nicholson <dbn.lists@gmail.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
ActivateDevice was ignoring errors from DeviceCursorInitialize, so
cursor-related calls failed later. Jeremy Huddleston saw that crash in
miPointerConstrainCursor, while with Xvfb I saw it in
miSpriteRealizeCursor.
miDCDeviceCleanup frees any non-NULL GCs. miDCDeviceInitialize calls
Cleanup on any failure, but if it failed early then some of the pointers
in the miDCBufferPtr were garbage. Switch from malloc to calloc to
ensure everything's initialized safely first.
With these two fixes, if CreateGC fails then the server gracefully fails
in FatalError instead of segfaulting.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Cc: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
As of e2929db7b7, doPolyText uses pFont
consistently rather than looking it up again from the saved XID.
clang noticed that "oldfid = fid" could run when fid hadn't been
initialized yet.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Jeremy Huddleston <jeremyhu@apple.com>
Tested-by: Jeremy Huddleston <jeremyhu@apple.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
The code this comment was referring to was removed in
8b5086250a "Eliminate bogus event resizing."
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
This eliminates a poorly-named, poorly-documented field from the
ScreenRec, using a previously-unused flag bit in each GC instead.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Cc: Keith Packard <keithp@keithp.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Rather than always returning BadValue, associate an error status like
BadWindow with a resource type like RT_WINDOW, and return the
appropriate one for the requested type.
This patch only touches the core protocol resource types. Others still
return BadValue and need to be mapped appropriately.
dixLookupResourceByType can now return BadImplementation, if the caller
asked for a resource type that has not been allocated in the server.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
If filter is NoEventMask (aka CantBeFiltered), grab is null, and the
first event is not in the set of "critical events", then TryClientEvents
simply calls WriteEventsToClient. In that case, it returns 0 for fake or
dead clients, and 1 otherwise. Inline for this special case.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Reviewed-by: Adam Jackson <ajax@redhat.com>
TryClientEvents already did this; this commit just moves the assignment
one level down so that no event source has to worry about sequence
numbers.
...No event source, that is, except XKB, which inexplicably calls
WriteToClient directly for several events.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Reviewed-by: Adam Jackson <ajax@redhat.com>
This matches the test in TryClientEvents, and is a superset of tests
done by the callers of these functions. The consequence of forgetting
these tests is a server crash, so they're always desirable. In my
opinion, it's better to not require the callers to remember to do these
checks.
For callers that don't do very much work before calling WriteToClient or
WriteEventsToClient, I've removed the redundant checks.
hw/xquartz/xpr/appledri.c has an interesting case: While its check for
"client == NULL" appears redundant with the test in WriteEventsToClient,
it dereferences client to get the sequence number.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=27497
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Reviewed-by: Adam Jackson <ajax@redhat.com>
dix/colormap.c and dix/gc.c now dereference a ClientPtr, so they need to
include dixstruct.h. Regression introduced by commit
11c69880c7.
Reported-by: Robert Hooker <sarvatt@ubuntu.com>
Signed-off-by: Julien Cristau <jcristau@debian.org>
Reviewed-by: Jamey Sharp <jamey@minilop.net>
Signed-off-by: Keith Packard <keithp@keithp.com>
Just let Dispatch() check for a noClientException, rather than making
every single dispatch procedure take care of it.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
In order to execute a wire-level ChangeGC request, we need to look up
the resources named by any XIDs in the value-list. Various places in the
server already have pointers to the resources they want to set into the
GC, though, so over time the interface has evolved to accept either XIDs
or pointers, with several different function call signatures used in
different eras.
This patch makes the existing code require pointers to resources rather
than XIDs, and adds a simple wrapper that looks up any XIDs. The old
dixChangeGC API is preserved by delegating to whichever implementation
is appropriate.
This affects error-handling: If any of the XIDs are invalid, then the GC
is unchanged, and its ChangeGC callback is not invoked. This change is
allowed by the protocol spec, which says, "The order in which components
are verified and altered is server-dependent. If an error is generated,
a subset of the components may have been altered."
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
XSELinux was the only consumer of these interfaces and it no longer
needs them.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jamey Sharp <jamey@minilop.net>
Otherwise we can't check that the XIDs this GC is being initialized with
are accessible to this client.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Cc: Eamon Walsh <ewalsh@tycho.nsa.gov>
Reviewed-by: Eamon Walsh <ewalsh@tycho.nsa.gov>
This doesn't change any behavior, but it isn't clear whether NullClient
is correct in all cases. As ajax says,
> For most of these changes, I think it's correct to use NullClient,
> since they are server-initiated changes and should not fail for (eg)
> xace reasons. ... At any rate, you're certainly not changing any
> semantics by leaving them all as NullClient, so this patch can't be
> more wrong than before.
The call in CreateGC is particularly questionable.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
This doesn't change any behavior, but it isn't clear whether NullClient
is correct in all cases. As ajax says,
> For most of these changes, I think it's correct to use NullClient,
> since they are server-initiated changes and should not fail for (eg)
> xace reasons. ... At any rate, you're certainly not changing any
> semantics by leaving them all as NullClient, so this patch can't be
> more wrong than before.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
In commit 42d6112ec2, Eamon changed
dixChangeGC to require DixUseAccess on any GCFont XID. I think
doPolyText needs to require the same level of access. Otherwise
dixChangeGC could fail when it does the same lookup, which doPolyText
doesn't check for.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Cc: Eamon Walsh <ewalsh@tycho.nsa.gov>
Reviewed-by: Eamon Walsh <ewalsh@tycho.nsa.gov>
Previously the callers were only setting errorValue on Success, when
it's ignored, and leaving it alone on failure, when it's sent to the
client.
Since SetFontPath takes the ClientPtr, let it set client->errorValue
instead of letting the callers continue to get it wrong.
Signed-off-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
Using one variant of function/macro makes it easier to fix the code
later.
Signed-off-by: Mikhail Gusarov <dottedmag@dottedmag.net>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
The only remaining X-functions used in server are XNF*, the rest is converted to
plain alloc/calloc/realloc/free/strdup.
X* functions are still exported from server and x* macros are still defined in
header file, so both ABI and API are not affected by this change.
Signed-off-by: Mikhail Gusarov <dottedmag@dottedmag.net>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
FreeResource() keeps clientTable[cid].elements up to date with the
number of resources allocated to the client. The other free
resource functions (FreeResourceByType(),
FreeClientNeverRetainResources() and FreeClientResources()) don't
maintain this invariant.
Typically, the only consequence is that the element count is too high
and we end up allocating the hash table bigger than necessary. However,
FreeResource() also relies on the element count to restart the search if
the list of resources has been changed during a resource destruction
callback. Since FreeResourceByType() doesn't update the count, if we call
that from a resource destruction callback from FreeResource(), the
loop isn't restarted and we end up following an invalid next pointer.
Furthermore, LookupClientResourceComplex() and
FreeClientNeverRetainResources() don't use the element count to detect
if a callback deleted a resource and may end up following an invalid
next pointer if the resource system is called into recursively.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
