This would let you do a constant-time context lookup, but if that's your
performance problem you have two problems. Just use the context's XID
as the tag value instead.
In order to do this, we have to defer destroying a context until it
actually goes unreferenced, as you're allowed to mention a context tag
after you've (ostensibly) destroyed the context, as long as it's still
your current context. Thus, change DestroyContext to merely mark the
context as dead if it's a current context, and call down to actual
resource destruction (and XID reclamation) in StopUsingContext.
Also, stop trying to delete context state from DrawableGone. This was
always broken, as GLX does not say that contexts are destroyed when
their drawables are destroyed. But with the above change to defer
context destruction, this would trigger a server crash on client exit as
we'd free the context state twice.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>
We never need to generate stubs, because those conditions can't happen
in the server. Yank that code out, but keep the bookkeeping for which
extension functions are registered so the DRI driver doesn't get
confused.
As a pleasant bonus, we're now friendlier for environments like selinux
that make runtime code generation difficult, and we're portable to more
arches since we don't have to port the assembly stubs.
Fixes the following clutter conformance tests (indirect rendering,
llvmpipe driver):
test-cogl-backface-culling
test-cogl-materials
test-cogl-readpixels
test-cogl-texture-mipmaps
test-cogl-texture-get-set-data
test-cogl-viewport
test-cogl-offscreen
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>
This isn't a meaningful thing in the indirect glx loader, so just warn
if it ever happens and move on.
But also, mark it PUBLIC, so if the driver does ever call it we merely
warn instead of aborting because ld.so can't find the symbol.
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>
We can never hit this, because the indirect GLX dispatch code always
forces a current context and checks that it's non-NULL before calling
into the dispatch table. If it's _not_ null, then _glapi_set_context
will call into the driver, which is responsible for calling
_glapi_set_dispatch to make sure the dispatch table is non-NULL.
Also remove _glapi_set_warning_func and friends, since we can no longer
call them even from dead code.
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>
DetermineClientPid didn't close file descriptor if read on
/proc/pid/cmdline failed. Adjusted the code to disregard the close
return value and perform the return after that, if the read failed or
returned EOF.
Signed-off-by: Mark Kettenis <mark.kettenis@xs4all.nl>
Signed-off-by: Erkki Seppälä <erkki.seppala@vincit.fi>
Reviewed-by: Rami Ylimäki <rami.ylimaki@vincit.fi>
Signed-off-by: Keith Packard <keithp@keithp.com>
This refers to pWin->winSize in some #if 0 code remove it.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Cyril Brulebois <kibi@debian.org>
each DDX has its own copy, I've taken the darwin one,
though I'm not sure why it needs the pOldClip piece that nobody
else has and the commit msg is like an "Updates from magic land"
type message.
This removes the main uses of pWin->winSize from the DDXen.
v2: drop old clip like ajax suggests.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
This removes the struct, but keeps InitAbsoluteClassDeviceStruct as
a no-op and preserves related struct layout.
Signed-off-by: Simon Thum <simon.thum@gmx.de>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Simon Thum <simon.thum@gmx.de>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Simon Thum <simon.thum@gmx.de>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Simon Thum <simon.thum@gmx.de>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Simon Thum <simon.thum@gmx.de>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Catch compiler errors that were otherwise only spotted on make check.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Dan Nicholson <dbn.lists@gmail.com>
wakeup_handler in udev.c wasn't dealing with udev change events.
There are situations when a device can gain its input capabilities
after it has been added to the system and therefore the change events
must be handled as well.
The change is handled as a consecutive device removal and addition.
Signed-off-by: Erkki Seppälä <erkki.seppala@vincit.fi>
Signed-off-by: Stefan Kost <Stefan.Kost@nokia.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Makes reading the log file a lot easier for those that don't magically
recognise the log spew by the individual drivers.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Dan Nicholson <dbn.lists@gmail.com>
AX_TLS detects when toolchains support __thread or __declspec(thread),
but existing code assumed __thread.
This also adds a check to configure.ac to error out if TLS is requested
but unsupported.
Found-by: Tinderbox
http://tinderbox.x.org/builds/2011-03-22-0007
Regression-from: 82b1eaa6ca
Signed-off-by: Jeremy Huddleston <jeremyhu@apple.com>
Reviewed-by: Tom Fogal <tfogal@alumni.unh.edu>
Reviewed-by: Dan Nicholson <dbn.lists@gmail.com>
A handful of modules have begun adding unit test programs.
These macros will help providing a consistent interface which will
help package builders and developers to manage the functionality.
XORG_ENABLE_UNIT_TESTS will turn on/off unit testing, regardless
of how it is implemented. The default (yes/no) can be specified by each
module. It can be used by itself if glib or -wrap support is not needed.
XORG_WITH_GLIB will probe the system for glib-2.0. A different version
can be specified in each module. It will consult XORG_ENABLE_UNIT_TESTS
but can be used by itself in contexts other then unit testing.
The default (yes/no) can be specified by each module.
XORG_LD_WRAP will probe the linker for -wrap support. It will consult
XORG_ENABLE_UNIT_TESTS but can be used by itself in contexts
other then unit testing.
configure options:
--enable-unit-tests Enable building unit test cases (default: auto)
--with-glib Use GLib library for unit testing (default: auto)
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Keith Packard <keithp@keithp.com>
No functional changes.
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Keith Packard <keithp@keithp.com>
This reverts commit 1564c82417.
The drivers used the top bits of the usage_hint to store driver
private flags (intel, radeon, nouveau).
With EXA we need to get at this data so if we migrate the pixmap we
can create the correct type of pixmap in the driver, however this
commit truncates the usage_hint into 8-bit class and loses all the
good stuff.
Signed-off-by: Dave Airlie <airlied@gmail.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Valgrind complains about uninitialized data being written to clients.
Reviewed-by: Erkki Seppälä <erkki.seppala@vincit.fi>
Signed-off-by: Rami Ylimäki <rami.ylimaki@vincit.fi>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This was leftover from some older ways of building dmx/scale docbook.
Reviewed-by: Dan Nicholson <dbn.lists@gmail.com>
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Keith Packard <keithp@keithp.com>
The convention is to have the manual pages in a man subdir
which is not under a doc dir. The doc dir contains users docs.
This will move man pages out of the way for upcoming DocBook patches.
Reviewed-by: Dan Nicholson <dbn.lists@gmail.com>
Signed-off-by: Gaetan Nadon <memsize@videotron.ca>
Signed-off-by: Keith Packard <keithp@keithp.com>
There were two memory leaks in the function: one was the lack of free
for "enabled", the other was the full lack of releasing anything when
configuration was too small. The first issue was fixed by adding the
missing free, the other was addressed by replacing the duplicate
memory releasing sequences with one that is gotoed into.
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Rami Ylimäki <rami.ylimaki@vincit.fi>
Signed-off-by: Erkki Seppälä <erkki.seppala@vincit.fi>
Signed-off-by: Keith Packard <keithp@keithp.com>
When XDMCP -from is specified, only register the requested address,
rather than the requested address, and any others we have of different
address families to the requested address.
e.g. if we have 4 interfaces with both IPv4 and IPv6 addresses (which
are not IPv6 mapped IPV4 addresses), using -from with one of those IPv4
addresses currently means only that IPv4 address, and all IPv6 addresses
are used in the connection data in XDMCP REQUEST packet.
(See http://cygwin.com/ml/cygwin-xfree/2011-02/msg00000.html)
Signed-off-by: Jon TURNEY <jon.turney@dronecode.org.uk>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
These functions no longer go through the screen vtable, so remove
them and fix up the various wrappers.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Acked-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Soren Sandmann <ssp@redhat.com>
There is no need to virtualize this function that nobody cares about.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Acked-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Soren Sandmann <ssp@redhat.com>
There is no need to virtualize this function that nobody cares about.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Acked-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Soren Sandmann <ssp@redhat.com>
The record allocated by miSpriteDeviceCursorInitialize was not being
released.
This patch makes misprite use dixRegisterPrivateKey with the record
size argument, which handles the memory management
issues. miSpriteDeviceCursorInitialize is restructured to initialize
pCursorInfo only if miDCDeviceInitialize succeeds. The record itself
is zeroed on cleanup to ensure that the assumptions in the code still
hold.
Reviewed-by: Rami Ylimäki <rami.ylimaki@vincit.fi>
Signed-off-by: Erkki Seppälä <erkki.seppala@vincit.fi>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
The record allocated by miSpriteDeviceCursorInitialize was not being
released.
This patch makes misprite use dixRegisterPrivateKey with the record
size argument, which handles the memory management issues.
miSpriteDeviceCursorInitialize is restructured to initialize pCursorInfo
only if miDCDeviceInitialize succeeds. The record itself is zeroed on
cleanup to ensure that the assumptions in the code still hold.
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Reviewed-by: Rami Ylimäki <rami.ylimaki@vincit.fi>
Signed-off-by: Erkki Seppälä <erkki.seppala@vincit.fi>
The only user of the geometry coordinates is the software sprite code,
which uses them to remove the pointer whenever the window beneath is
being used as a source. However, using Window pictures as a source is
extremely rare (let alone *partial* windows), so there is no harm done
in just validating all of the drawable.
Additionally, the miSourceValidate() function was buggy in at least
three respects:
(a) It added drawable->{x,y} before calling down, which is wrong since
the misprite code already adds them in its check. (Alternatively,
the misprite code is wrong, but there are actual users who would
notice if that code was broken).
(b) It didn't account for the width of the interpolation filter, so if
the Picture had a bilinear or convolution filter, the edges
surrounding the source area would not be validated.
(c) It didn't validate alpha maps.
Finally, computing the bounding box of the transform on every
composite request was a real performance issue in pixman, so
presumably it could be one here as well.
This patch changes miSourceValidate() to simply validate all of the
underlying drawable.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Soren Sandmann <ssp@redhat.com>
long is needlessly long on LP64.
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
size needn't be a long. No change on ILP32 but, combined with the
previous change, 56 -> 40 bytes on LP64.
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
unsigned long is needlessly large on LP64. Use uint32_t instead.
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
DRI, DRI2 and swrast all had near-identical driver probing logic.
Pull it into glxdricommon.
[ajax: warning fix]
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Christopher James Halse Rogers <christopher.halse.rogers@canonical.com>
Confine cursor motion to within the bounds of a single CRTC, iff all the
CRTCs within a ScreenRec are reachable from each other. If not you get
the same "cursor floats within the bounding rect" behaviour you get now.
v3:
- Incorporate review feedback from Christopher James Halse Rogers
v4:
- Add mode field.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
