A few other variables are unused but they are setting using register
reads so I am afraid to touch them.
Signed-off-by: Connor Behan <connor.behan@gmail.com>
On 64bit the long int type is 64 bit, but the MSR operations still
are 32bit, thus we need to use uint32_t. Also need slighty modified
versions of the inline asm code, since on 64bit machines, the 32bit
(pseudo-)registers can't be used on push/pop.
This patch is mostly for CI purpose, since there isn't any known version
of this graphics device on a 64 bit machine.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
In addition to removing some more dead code, this warns the user that
OSMImageBuffers now does nothing. The similar option OSMColorExpBuffers
is still used by EXA so there is no warning for this.
Signed-off-by: Connor Behan <connor.behan@gmail.com>
One of the checks for HAVE_GX appeared to be in the wrong place. Once it
is in the right place, a configure option can be used to not set HAVE_GX
and avoid compiling the associated files.
Signed-off-by: Connor Behan <connor.behan@gmail.com>
When the minimum Xserver version was recently bumped to 1.18, the
ability to enable XAA was removed accordingly. But plenty of XAA hooks
were left in. Removing these will lighten the code and get rid of some
more compiler warnings.
Signed-off-by: Connor Behan <connor.behan@gmail.com>
Long ago, xf86_reload_cursors() was turned into a noop. And we were not
using anything specific from xf86PciInfo.h.
Signed-off-by: Connor Behan <connor.behan@gmail.com>
This applies some obvious changes to stop gcc from complaining about
dead code, shadow declarations, differing signedness and sections that
mix declarations with code.
The warning about discarding const qualifiers is more annoying because
we pass string literals to some functions which accept non-const
pointers. Currently, this takes the following approach. If the function
*needs* to accept non-const pointers, cast the string literal as char *.
Otherwise, change the function to only accept a const pointer. To
anticipate future use cases though, I could also leave function
definitions as they are and just always cast string literals.
Alternatively, if this is more trouble that it is worth, we could just
put up with the warnings.
Signed-off-by: Connor Behan <connor.behan@gmail.com>
There's used to have separate versions for older Xserver versions, but no
anymore, so these aren't needed anymore.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
Since recent commits require xserver-1.18.0 or later to build against,
there's no reason leaving behind big chunks of code that can only build
against the XAA support removed in xserver-1.13.0 (released in 2012).
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
Causes autoreconf to fail with:
libtoolize: error: AC_CONFIG_MACRO_DIRS([m4]) conflicts with ACLOCAL_AMFLAGS
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Thanks to Ben Hutchings for pointing out that vesafb and legacy X
drivers can peacefully coexist if iomem=relaxed gets added to the
Linux kernel cmdline options.
Signed-off-by: Martin-Éric Racine <martin-eric.racine@iki.fi>
This release restores compatibility with xserver-1.20 and newer.
VALIDATION PLATFORM
* Debian (X server 1.20.4) on FIC ION603A (Geode LX800).
Signed-off-by: Martin-Éric Racine <martin-eric.racine@iki.fi>
xf86DisableRandR() doesn't exist anymore, and we don't need it anyway,
the core code will notice that we set up RANDR ourselves.
Signed-off-by: Adam Jackson <ajax@redhat.com>
This release restores compatibility with xserver-1.19 and newer.
VALIDATION PLATFORM
* Debian (X server 1.19.0) on FIC ION603A (Geode LX800).
Signed-off-by: Martin-Éric Racine <martin-eric.racine@iki.fi>
