Emil Velikov writes:
> This will enable it on more platforms than just Linux.
Right, the goal is to use it where available.
> You can use AX_PTHREAD here, instead of open-coding it.
Thanks. I've updated the series to do this; it's on my input-thread
branch in
AC_DEFAULT_BUNDLE_VERSION for the m4 variable avoids this
problem. At the same time, remove an extra newline in the m4 value.
Signed-off-by: Keith Packard <kei...@keithp.com>
---
configure.ac | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/configure.ac b/configure.ac
Signed-off-by: Keith Packard <kei...@keithp.com>
Acked-by: Peter Hutterer <peter.hutte...@who-t.net>
---
hw/kdrive/src/kinput.c | 6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/hw/kdrive/src/kinput.c b/hw/kdrive/src/kinput.c
index b415a86..4119b08 100644
---
Now that events are read at normal process time, we can use malloc to
grow the event queue instead of discarding events.
Signed-off-by: Keith Packard <kei...@keithp.com>
---
mi/mieq.c | 103 +++---
1 file changed, 32 insertions(
Now that input is threaded, malloc can be used at event time to resize
the touchpoint buffer as needed.x
Signed-off-by: Keith Packard <kei...@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutte...@who-t.net>
---
dix/touch.c | 89 +---
that expose the same API/ABI.
Respond to style comments from Peter Hutterer.
v5: use __func__ in inputthread debug and error mesages.
Respond to style comments from Peter Hutterer.
Signed-off-by: Adam Jackson <a...@redhat.com>
Signed-off-by: Keith Packard <kei...@keithp.c
When this code was called from SIGIO, saving and restoring errno could
possibly have made sense in some strange environment. Now that this
will not be called from a signal handler, there is no reason to do that.
Signed-off-by: Keith Packard <kei...@keithp.com>
---
hw/xfree86/common/xf86Ev
,
xf86CurrentCursor, which returns the current cursor. With this new
function, we can eliminate the cursor pointer from the
xf86CrtcConfigRec, once drivers are converted over to use it.
Signed-off-by: Keith Packard <kei...@keithp.com>
---
hw/xfree86/modes/xf86Cursors.c | 17 +
hw/xfree86/
Switch the XFree86 DDX over to threaded input
v2: Rewrite comment in xf86Helper about silken mouse
Signed-off-by: Keith Packard <kei...@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutte...@who-t.net>
---
hw/xfree86/common/xf86Events.c | 4 ++--
hw/xfree86/common/xf86H
Signed-off-by: Keith Packard <kei...@keithp.com>
---
Xi/exevents.c | 4 +-
config/config.c | 4 +-
configure.ac | 14 ---
dix/devices.c | 10 +-
This makes the code more consistent with other versions of
out-of-queue event processing
Signed-off-by: Keith Packard <kei...@keithp.com>
---
xkb/xkbActions.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/xkb/xkbActions.c b/xkb/xkbActions.c
index afe5edf..048ed44
,
xf86CurrentCursor, which returns the current cursor. With this new
function, we can eliminate the cursor pointer from the
xf86CrtcConfigRec, once drivers are converted over to use it.
Signed-off-by: Keith Packard <kei...@keithp.com>
---
hw/xfree86/modes/xf86Cursors.c | 17 +
hw/xfree86/
t was supposed
to, in which case vblank->target_msc will be exactly crtc_msc+1, or it's
being done late, in which case that will have passed and we'll copy it
immediately.
Reviewed-by: Keith Packard <kei...@keithp.com>
--
-keith
signature.asc
Description: PGP signature
_
mageCheck function doesn't get the CursorPtr. So, we
need the xf86CurrentCursor function that I had written before.
From c133e4d9fac39862d7922f59080cc7ce40d1108b Mon Sep 17 00:00:00 2001
From: Keith Packard <kei...@keithp.com>
Date: Wed, 9 Mar 2016 11:13:14 -0800
Subject: [PATCH xserver] xfree86:
v2: Check crtc->transformPresent instead of crtc->transform_in_use. The
> latter is TRUE for rotation as well, which we handle correctly.
> Reviewed-by: Keith Packard <kei...@keithp.com>
--
-keith
signature.asc
Description: PGP signature
__
Michel Dänzer writes:
> Unfortunately, the reality is that at least the radeon and amdgpu
> drivers do use it.
Even modesetting uses it; there isn't any other public interface to get
the current cursor.
What do you think about eliminating the RefCursor/FreeCursor bits and
Michel Dänzer <mic...@daenzer.net> writes:
> From: Michel Dänzer <michel.daen...@amd.com>
>
> This reduces code duplication.
>
> v2: No functional change this time.
> Signed-off-by: Michel Dänzer <michel.daen...@amd.com>
Reviewed-by: Keith Packard <kei...@
Michel Dänzer writes:
> It seems to work fine in my quick testing. However, I'm not sure it's
> worth it, given that v2 of patch 1 fixes holding on to the last HW
> cursor indefinitely, and given the churn it would cause for external
> drivers.
I guess I don't see how this
Dave Airlie <airl...@gmail.com> writes:
> Signed-off-by: Dave Airlie <airl...@redhat.com>
Reviewed-by: Keith Packard <kei...@keithp.com>
--
-keith
signature.asc
Description: PGP signature
___
xorg-devel@lists.x.org: X.Org devel
le of
times to make sure the outputs all get repainted, and that should end up
calling ClipNotify on any flip window. So,
Reviewed-by: Keith Packard <kei...@keithp.com>
--
-keith
signature.asc
Description: PGP signature
___
xorg-devel@lists.x.org: X.Org
s slightly simplified, exploiting
>> the fact that this function is only ever called with a valid flip
>> pixmap.
>>
>> v2: Assert that flip_pixmap is non-NULL instead of testing and bailing
>> on NULL, preserve test for flip_window being non-NULL before calling
>> pr
eady the case in other present code.
>
> Signed-off-by: Michel Dänzer <michel.daen...@amd.com>
Reviewed-by: Keith Packard <kei...@keithp.com>
--
-keith
signature.asc
Description: PGP signature
___
xorg-devel@lists.x.org: X.Org dev
Michel Dänzer writes:
> From: Michel Dänzer
>
> The following fix will use the refactored function.
This isn't a simple refactor; the semantics of the new function are
different from the replaced code. I'd love to see a separate patch which
changed
n switching a flipping window out of fullscreen, a repeated unflip
> copy could leave artifacts of the flip pixmap on the desktop.
>
> Signed-off-by: Michel Dänzer <michel.daen...@amd.com>
Reviewed-by: Keith Packard <kei...@keithp.com>
> +/* Update the screen pixmap wi
Michel Dänzer writes:
Here's another review, now that I've more carefully looked at the other
two patches.
> +if (!flip_window || !flip_pixmap)
> +return;
In the original code, this didn't abort the operation entirely...
This operation would have been skipped:
Dave Airlie writes:
> Seems like a good thing to do.
>
> Reviewed-by: Dave Airlie
Merged.
--
-keith
signature.asc
Description: PGP signature
___
xorg-devel@lists.x.org: X.Org development
Archives:
This lets clients know that the layout of the monitors on the screen
has changed so they can adapt appropriately.
Signed-off-by: Keith Packard <kei...@keithp.com>
---
randr/rrmonitor.c | 9 +++--
1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/randr/rrmonitor.c b
Adam Jackson writes:
> I suspect this code predates the common resource hooks for computing
> sizes. It's ugly in any case since the Resource extension shouldn't
> need to know which extensions can take a reference on pixmaps. Instead,
> let's just walk every resource for the
Adam Jackson writes:
> I don't really understand your objection about pixmaps not in the
> resdb, because ResQueryClientPixmapBytes is _already_ not accounting
> for those, and it's not clear to me how it even could.
Sorry, that code has gotten a lot more sophisticated since I
Michel Dänzer writes:
> I'm not sure that makes sense; e.g. it seems inconsistent with leaving
> the cursor image untransformed in the driverIsPerformingTransform case.
> It seems to me like the idea behind driverIsPerformingTransform was to
> leave all transformations to the
Aaron Plattner <aplatt...@nvidia.com> writes:
> On 01/29/2016 07:23 AM, Keith Packard wrote:
>> Michel Dänzer <mic...@daenzer.net> writes:
>>
>>> I'm not sure that makes sense; e.g. it seems inconsistent with leaving
>>> the cursor image untransf
gt;
> v2: Return BadMatch instead of BadRequest (Keith Packard)
> Signed-off-by: Michel Dänzer <michel.daen...@amd.com>
Reviewed-by: Keith Packard <kei...@keithp.com>
(might want to update the DRI3 spec to reflect this change?)
--
Eric Anholt writes:
> Yes, it's supposed to, but we don't actually store 16bpp anything in
> glamor, because we suck.
Ouch.
--
-keith
signature.asc
Description: PGP signature
___
xorg-devel@lists.x.org: X.Org development
Archives:
Eric Anholt writes:
> if (op == PictOpSrc) {
> +/* We can't do direct copies between different depths at 16bpp
> + * because r,g,b are allocated to different bits.
> + */
> +if (dst->pDrawable->bitsPerPixel == 16 &&
> +
gt; +}
> +
These should probably get declared as inline, although the compiler will
probably just inline them anyways.
I also won't be surprised if we end needing to expose these in a header
for other code to use too. Should we just do that now so that someone
needing to compare values find
heck for that as well then, just in case the
driver has some other crazy error states.
Sorry for not adding my
Reviewed-by: Keith Packard <kei...@keithp.com>
before!
--
-keith
signature.asc
Description: PGP signature
___
xorg-devel@lists.x.org:
Dave Airlie writes:
> From: Dave Airlie
>
> running xfontsel on haswell here, with a max texture size
> of 8kx8k, one font wants 9711 height. This fallsback to
> sw in this case.
>
> A proper solution probably involves using an array texture.
Given that
> tx = (col - firstCol) * glyph_spacing_x;
> +/* adjust for second row layout */
> +tx += second_row * glamor_font->row_width * 8;
I'm not OK using a Bool for multiplication here. Just declare it as an
int and initialize to zero and we'll be good.
Otherwi
Dave Airlie writes:
> From: Dave Airlie
>
> running xfontsel on haswell here, with a max texture size
> of 8kx8k, one font wants 9711 height. This fallsback to
> sw in this case.
>
> A proper solution probably involves using an array texture.
>
>
Adam Jackson writes:
> Leo's patch to fix the EDID atom leak is probably correct, in the sense that
> it fixes the leak, but xf86RegisterRootWindowProperty is fairly ugly to
> begin with so I'd rather see it eliminated. This series adds a generic
> callback for screen
Adam Jackson writes:
> Oops. This turns out to cause a crash on init, I've pushed a revert
> until I figure out why.
The patch turns one statement into two; presumably that matters
somewhere.
--
-keith
signature.asc
Description: PGP signature
TransConns entries before the point that ListenTransFds
> was allocated & initialized, so CloseWellKnownConnections would
> segfault trying to read entries from a NULL ListenTransFds pointer.
>
> Introduced by commit 7b02f0b8
>
> Signed-off-by: Alan Coopersmith <alan.coopersm...@o
I finally dug a 4k monitors out and plugged it into my laptop. Of
course, my laptop needs to use MST to drive it at 60Hz.
I got to fix version 2.24 of gtk+:
https://bugzilla.gnome.org/show_bug.cgi?id=759912
Next, I wanted to automatically configure things when I plugged and
unplugged
Michel Dänzer writes:
> From: Michel Dänzer
>
> Even if the driver is handling the transform, we still need to transform
> the cursor position for clipping, otherwise we may hide the HW cursor
> when the cursor is actually inside the area covered by
Michel Dänzer <mic...@daenzer.net> writes:
> From: Michel Dänzer <michel.daen...@amd.com>
Yeah, looks cleaner this way.
Reviewed-by: Keith Packard <kei...@keithp.com>
--
-keith
signature.asc
Description: PGP signature
___
x
Michel Dänzer <mic...@daenzer.net> writes:
> On 26.12.2015 04:08, Keith Packard wrote:
>> Michel Dänzer <mic...@daenzer.net> writes:
>>
>>> From: Michel Dänzer <michel.daen...@amd.com>
>>>
>>> Even if the driver is handling the t
anymore.
>
> Signed-off-by: Michel Dänzer <michel.daen...@amd.com>
Reviewed-by: Keith Packard <kei...@keithp.com>
--
-keith
signature.asc
Description: PGP signature
___
xorg-devel@lists.x.org: X.Org development
Archives: http://list
Michel Dänzer writes:
> From: Michel Dänzer
>
> It was already disabled, but it's definitely no longer needed now that
> xf86CursorResetCursor is getting called for each CRTC configuration
> change.
I think we can remove xf86_reload_cursors from the
.2 cursor code handles transforms correctly, we can
> allow them to properly take effect.
>
> Signed-off-by: Michel Dänzer <michel.daen...@amd.com>
I've already marked this as:
Reviewed-by: Keith Packard <kei...@keithp.com>
--
-keith
signature.asc
Description: PGP signature
Michel Dänzer <mic...@daenzer.net> writes:
> From: Michel Dänzer <michel.daen...@amd.com>
>
> We currently don't handle transforms for the HW cursor image, so return
> FALSE to signal a software cursor must be used if a transform is in use
> on any CRTC.
Revi
ld' work.
From da2e5c4f6e08b765a6a3e7337c0da8dd60b531e0 Mon Sep 17 00:00:00 2001
From: Keith Packard <kei...@keithp.com>
Date: Thu, 24 Dec 2015 09:51:49 -0800
Subject: [PATCH xserver] xfree86/modes: Remove extra reference to current
cursor
Just use the cursor reference in the xf86CursorScree
s this right above the call to
set_mode_major, making this change "obviously" correct).
Reviewed-by: Keith Packard <kei...@keithp.com>
--
-keith
signature.asc
Description: PGP signature
___
xorg-devel@lists.x.org: X.Org development
Archi
Michel Dänzer <mic...@daenzer.net> writes:
> From: Michel Dänzer <michel.daen...@amd.com>
>
> This will disable the HW cursor while a transform is active on the CRTC.
>
> Signed-off-by: Michel Dänzer <michel.daen...@amd.com>
Reviewed-by: Keith Packar
Michel Dänzer <mic...@daenzer.net> writes:
> From: Michel Dänzer <michel.daen...@amd.com>
>
> This allows switching between HW and SW cursor depending on the current
> state.
Reviewed-by: Keith Packard <kei...@keithp.com>
--
-keith
signature.
Michel Dänzer writes:
> From: Michel Dänzer
>
> This will switch from HW cursor to SW cursor in the following cases:
> * A transform is present on a CRTC
> * This screen is a GPU screen acting as an output slave for another
> screen
Could this be
Michel Dänzer writes:
> From: Michel Dänzer
>
> Allows drivers to report whether or not the HW cursor can be used on a
> per-CRTC basis. If the driver reports that the HW cursor cannot be used
> on any CRTC, SW cursor has to be used for the screen.
>
Michel Dänzer writes:
> New series addressing this on the way. Thanks for your review of this
> series.
Cool! Love it when review makes things better.
--
-keith
signature.asc
Description: PGP signature
___
Michel Dänzer writes:
> Without this patch, there's no way for a driver using the code in
> xf86Cursors.c to force an SW cursor.
Ok, sorry I was just being dense and not understanding what the code was
actually doing. Thanks for the explanation, I understand the aim of
the
Michel Dänzer writes:
> This series addresses two cases in which a HW cursor cannot be used:
>
> 1) When a transform is active on a CRTC. Patch 2 adds a CRTC hook
>use_hw_cursor_argb, which allows drivers to force SW cursor on a
>per-CRTC basis. However, this hook
Michel Dänzer writes:
> This series addresses two cases in which a HW cursor cannot be used:
>
> 1) When a transform is active on a CRTC. Patch 2 adds a CRTC hook
>use_hw_cursor_argb, which allows drivers to force SW cursor on a
>per-CRTC basis. However, this hook
Michel Dänzer writes:
> 1) When a transform is active on a CRTC.
We haven't disabled the HW cursor when transforms are active in the
past, although I can imagine it'd be a nice thing to do.
--
-keith
signature.asc
Description: PGP signature
* Use silken mouse if requested and if we have threaded input
*/
pScrn->silkenMouse = useSM && InputThreadEnable;
if (serverGeneration == 1)
From 399961d249d8483f261824beea9a3a2b0eda97b1 Mon Sep 17 00:00:00 2001
From: Keith Packard <kei...@keithp.com>
Date: Tue, 8 Dec
Peter Hutterer writes:
> we should really move that lot into a helper function...
Well, now that growing the queue happens while queuing events, we might
just remove it entirely. Unless we run out of memory, it'll never happen
now.
>> -
>> -return;
>> +
driver '%s' for '%s'\n", drv->driverName,
> +pInfo->name);
Probably don't want to move this message; if systemd pauses things,
you'll get this message even if the device doesn't end up getting used.
Otherwise, this lgtm.
Reviewed-by: Keith Packard <kei...@ke
e list
> pointer comparison.
Done.
> __func__ for all three?
I don't understand; do you want __func__ to be displayed in these error
and debug printfs?
> please wrap the xorg_list loop in { }
Done.
Here's a new patch which merges 4/9 and 5/9 into 4/8
From 87d71e36185f2e1d7bc085c3f45
Peter Hutterer writes:
> typo: uppercase 't'
"fixed" by rewriting the commit message.
> seems a bit odd to use a case statement for a basic check. can we use a
> simple if here? (ignore this if a later patch changes this bit)
Sorry, old Bourne shell habit. Fixed. I
Now that input is threaded, malloc can be used at event time to resize
the touchpoint buffer as needed.x
---
dix/touch.c | 89 +
1 file changed, 30 insertions(+), 59 deletions(-)
diff --git a/dix/touch.c b/dix/touch.c
index
Signed-off-by: Keith Packard <kei...@keithp.com>
---
hw/kdrive/src/kinput.c | 6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/hw/kdrive/src/kinput.c b/hw/kdrive/src/kinput.c
index 8527a9e..446883b 100644
--- a/hw/kdrive/src/kinput.c
+++ b/hw/kdrive/src/kinput.c
@@
Now that events are read at normal process time, we can use malloc to
grow the event queue instead of discarding events.
Signed-off-by: Keith Packard <kei...@keithp.com>
---
mi/mieq.c | 103 +++---
1 file changed, 32 insertions(
e by default on non-Linux
This also splits out the actual enabling of input threads to
DDX-specific patches which follow
v3: Make the input lock recursive
Signed-off-by: Adam Jackson <a...@redhat.com>
Signed-off-by: Keith Packard <kei...@keithp.com>
recursive lock
---
configure.ac
---
include/input.h | 6 ++
1 file changed, 6 insertions(+)
diff --git a/include/input.h b/include/input.h
index d8bd9c6..5bc60f0 100644
--- a/include/input.h
+++ b/include/input.h
@@ -712,4 +712,10 @@ extern _X_HIDDEN void input_constrain_cursor(DeviceIntPtr
pDev, ScreenPtr screen
This makes the code more consistent with other versions of
out-of-queue event processing
Signed-off-by: Keith Packard <kei...@keithp.com>
---
xkb/xkbActions.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/xkb/xkbActions.c b/xkb/xkbActions.c
index afe5edf..048ed44
this
information.
xfree86 SIGIO support is reworked to use internal versions of
OsBlockSIGIO and OsReleaseSIGIO.
v2: Don't change locking order (Peter Hutterer)
Signed-off-by: Keith Packard <kei...@keithp.com>
---
Xi/exevents.c | 4 +-
config/co
Switch the XFree86 DDX over to threaded input
Signed-off-by: Keith Packard <kei...@keithp.com>
---
hw/xfree86/common/xf86Events.c | 4 ++--
hw/xfree86/common/xf86Helper.c | 2 +-
hw/xfree86/common/xf86Init.c | 3 +++
3 files changed, 6 insertions(+), 3 deletions(-)
diff --git a/hw/x
mutex support in the server. We prefer
PTHREAD_RECURSIVE_MUTEX_SUPPORt where available, falling back to
__thread/__declspec(thread) support if necessary.
Signed-off-by: Keith Packard <kei...@keithp.com>
fixup for thread variables
---
configure.ac| 14 ++
hw/xfree86/sdks
Peter Hutterer <peter.hutte...@who-t.net> writes:
> fd = xf86OpenSerial(pInfo->options);
> + xf86FlushInput(fd);
Reviewed-by: Keith Packard <kei...@keithp.com>
--
-keith
signature.asc
Description: PGP signature
___
x
argv)
{
int t;
pthread_t threads[THREADS];
for (t = 0; t < THREADS; t++)
pthread_create([t], NULL, doit, NULL);
for (t = 0; t < THREADS; t++)
pthread_join(threads[t], NULL);
}
From 55f483cff8d660e444a65026e57f793c06a66ce3 Mon Sep 17 00:00:00 2001
From: Keith Packard <kei...@keithp.com>
f
conditional code.
Here's code which works only with recursive mutexes, initializing it
explicitly when PTHREAD_RECURSIVE_MUTEX_INITIALIZER_NP is not available.
From 700fb83a0afbfce19bb91967f619cde6ead900a6 Mon Sep 17 00:00:00 2001
From: Keith Packard <kei...@keithp.com>
Date: Thu, 10 Dec 2015 11:3
Mark Kettenis writes:
> However, is there a reason why you didn't use the
> PTHREAD_MUTEX_RECURSIVE mtex type that is standardized by POSIX?
Sorry, my documentation only mentions a non-portable version of
recursive mutexes. I didn't even know that recursive mutexes had
Keith Packard <kei...@keithp.com> writes:
> Mark Kettenis <mark.kette...@xs4all.nl> writes:
>
>> However, is there a reason why you didn't use the
>> PTHREAD_MUTEX_RECURSIVE mtex type that is standardized by POSIX?
>
> Sorry, my documentation only mentions a
Mark Kettenis writes:
> Ugh. Exporting global variables as part of the ABI is generally not
> such a good idea. Perhaps it is better to use functions to acquire
> and release the input mutex instead?
Yeah, it's not a performance sensitive operation, so hiding the
Peter Hutterer writes:
> in light of Mark's comments to 4/8 these should be an opaque API rather than
> inlined so we don't need to export the globals.
Yeah, probably a good idea.a
--
-keith
signature.asc
Description: PGP signature
fying that if I misspell '__thread' as '__tread' it fails as
expected).
Here's just the diff from the existing patch; I'll squash this in if it
looks reasonable.
From ef34d763145deb636c771b9c7bd3fb1bcbaa08ad Mon Sep 17 00:00:00 2001
From: Keith Packard <kei...@keithp.com>
Date: Thu, 10 Dec 2015
Michel Dänzer writes:
> +if (!client->local)
> +return BadRequest;
BadRequest is probably not the error code you want here; perhaps
BadMatch?
--
-keith
signature.asc
Description: PGP signature
___
Peter Hutterer writes:
>> /* Call PIE here so we don't try to dereference a device that's
>> * already been removed. */
>> -OsBlockSignals();
>> ProcessInputEvents();
>> +input_lock();
>
> this is a behaviour change, was this intended? I'd rather
Matt Turner writes:
> ---
> Because two Makefiles are better than ten.
Seems like a fine change; I wonder why you bothered?
--
-keith
signature.asc
Description: PGP signature
___
xorg-devel@lists.x.org: X.Org development
Emil Velikov writes:
>> -BOOL serverRunning = FALSE;
>> +Bool serverRunning;
> Deliberate change ? The declaration in the headers is still using BOOL.
Oops. So many edits; I thought for a while that we'd want to use this
code to make the input thread wait for the
Peter Hutterer writes:
> this is a behaviour change, was this intended? I'd rather not have this
> hidden in a giant patch that is otherwise mostly search and replace.
Ok, I've pushed out a new series which creates a recursive (counting)
mutex that makes all of this
---
include/input.h | 6 ++
1 file changed, 6 insertions(+)
diff --git a/include/input.h b/include/input.h
index d8bd9c6..5bc60f0 100644
--- a/include/input.h
+++ b/include/input.h
@@ -712,4 +712,10 @@ extern _X_HIDDEN void input_constrain_cursor(DeviceIntPtr
pDev, ScreenPtr screen
this
information.
xfree86 SIGIO support is reworked to use internal versions of
OsBlockSIGIO and OsReleaseSIGIO.
v2: Don't change locking order (Peter Hutterer)
Signed-off-by: Keith Packard <kei...@keithp.com>
---
Xi/exevents.c | 4 +-
config/co
e by default on non-Linux
This also splits out the actual enabling of input threads to
DDX-specific patches which follow
v3: Make the input lock recursive
Signed-off-by: Adam Jackson <a...@redhat.com>
Signed-off-by: Keith Packard <kei...@keithp.com>
recursive lock
---
configure.ac
Now that events are read at normal process time, we can use malloc to
grow the event queue instead of discarding events.
Signed-off-by: Keith Packard <kei...@keithp.com>
---
mi/mieq.c | 103 +++---
1 file changed, 32 insertions(
Now that input is threaded, malloc can be used at event time to resize
the touchpoint buffer as needed.x
---
dix/touch.c | 89 +
1 file changed, 30 insertions(+), 59 deletions(-)
diff --git a/dix/touch.c b/dix/touch.c
index
This makes the code more consistent with other versions of
out-of-queue event processing
Signed-off-by: Keith Packard <kei...@keithp.com>
---
xkb/xkbActions.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/xkb/xkbActions.c b/xkb/xkbActions.c
index afe5edf..048ed44
Switch the XFree86 DDX over to threaded input
Signed-off-by: Keith Packard <kei...@keithp.com>
---
hw/xfree86/common/xf86Events.c | 4 ++--
hw/xfree86/common/xf86Helper.c | 2 +-
hw/xfree86/common/xf86Init.c | 3 +++
3 files changed, 6 insertions(+), 3 deletions(-)
diff --git a/hw/x
This is the second version of this series, the big change it to switch
to a counting mutex using a thread-local variable. This makes the
existing input_lock/input_unlock stuff 'just work' as the old
OsBlockSIGIO stuff had a counter too. There's one change in locking;
injected xkb events were being
Signed-off-by: Keith Packard <kei...@keithp.com>
---
hw/kdrive/src/kinput.c | 6 --
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/hw/kdrive/src/kinput.c b/hw/kdrive/src/kinput.c
index 8527a9e..446883b 100644
--- a/hw/kdrive/src/kinput.c
+++ b/hw/kdrive/src/kinput.c
@@
Mark Kettenis writes:
> Most of my objections to this Tiago's patch still apply. Threads
> severely impact debuggability and this needs a full audit of all the
> drivers; not just the server.
I added the change which makes -dumbSched disable input threading to
ease
Peter Hutterer writes:
> this hunk took me a while, wouldn't it be easier do do something like:
> do {
>for (i = 0, ...) {
>...
>}
>
>if (!ti)
>TouchResizeQueue(dev))
> } while (!ti);
It's similar, but you still
Here's a series which revives Tiago's ancient threaded input
patch. This has been cleaned up to make sure that it does locking
correctly, and then patches for xf86 and kdrive/fbdev hook the code to
a couple of X servers to show how it should work.
Before applying that, all of the SIGIO support in
801 - 900 of 4633 matches
Mail list logo