xf86-video-cirrus: Changes to 'master'

[Adam Jackson]

 src/Makefile.am  |   24 ++++++------------------
 src/alp_driver.c |   14 +-------------
 src/cir_driver.c |   43 ++-----------------------------------------
 src/lg_driver.c  |   15 +--------------
 4 files changed, 10 insertions(+), 86 deletions(-)

New commits:
commit 27f85fc523102f89c3fc1546ae500816ccbffc8a
Author: Adam Jackson <ajax at redhat.com>
Date:   Wed Apr 13 17:05:55 2016 -0400

    Don't build split alpine and laguna support
    
    Before:
    
       text	   data	    bss	    dec	    hex	filename
      25772	   2040	     16	  27828	   6cb4	src/.libs/cirrus_alpine.so
      11429	   1552	    112	  13093	   3325	src/.libs/cirrus_drv.so
      21968	   2016	     16	  24000	   5dc0	src/.libs/cirrus_laguna.so
      59169	   5608	    114	  64921	   fd99	(TOTALS)
    
    After:
    
       text	   data	    bss	    dec	    hex	filename
      47831	   3840	    112	  51783	   ca47	src/.libs/cirrus_drv.so
    
    So, in the worst case, you're using roughly two more pages for (clean,
    evictable) text, but two fewer (dirty) pages for data and bss. Fixing
    this also allows the X server to clean up module loading a bit by
    enforcing the _drv.so suffix, and since this is the only driver pulling
    such shenanigans...
    
    Reviewed-by: Julien Cristau <jcristau at debian.org>
    Signed-off-by: Adam Jackson <ajax at redhat.com>

commit adaf1fe744df8d37acdb8b157abc5034b8ea1073
Author: Adam Jackson <ajax at redhat.com>
Date:   Wed Apr 13 16:37:38 2016 -0400

    Remove (almost) no-op setup functions
    
    These became practical no-ops when I removed reference to the loader
    symbol lists. gcc will still emit code (and bss) for them though. No
    functional change, just doing it to prove that nothing special happens
    at submodule load.
    
    Reviewed-by: Julien Cristau <jcristau at debian.org>
    Signed-off-by: Adam Jackson <ajax at redhat.com>