xserver: Branch 'server-1.14-branch' - 10 commits

Aaron Plattner aplattner at nvidia.com
Mon Nov 4 01:10:52 CET 2013


On 13-11-02 07:44 PM, Matt Dew wrote:
> Hi Aaron,
>    Ugh.  Unacceptable for a stable branch.
>
> How do I fix this?
> Should I revert that whole merge?

I think it should be okay to just move the new fields to the end of the 
structures.

_rrProvider is calloc'd in RRProviderCreate.  It's immediately followed 
by the name of the provider, but RRProviderCreate stashes a pointer to 
the name in a field in the structure itself:

   provider->name = (char *) (provider + 1);

so as long as no one is trying to find the name by using (provider+1), 
that should be fine.

_rrScrPriv is calloc'd in RRScreenInit and stashed as a screen private 
using dixSetPrivate, so here again, adding to the end of the structure 
should be fine.

I can try to put together a patch and test it this week.  Is this an 
acceptable exception to the "only cherry-picks from master" policy for 
the stable branches?

> thanks,
> Matt
>
> On 11/01/2013 09:33 AM, Aaron Plattner wrote:
>> On 10/31/2013 11:03 PM, Matt Dew wrote:
>>>    hw/xfree86/common/xf86platformBus.c |    6 +
>>>    hw/xfree86/modes/xf86RandR12.c      |   21 +++++-
>>>    randr/randr.c                       |  120
>>> +++++++++++++++++++++++++++++++++++-
>>>    randr/randrstr.h                    |   12 +++
>>
>> These structures are part of the ABI, so these changes are going to fork
>> the ABI in an incompatible way from master.
>>
>>>    randr/rrcrtc.c                      |    6 +
>>>    randr/rrinfo.c                      |    2
>>>    randr/rroutput.c                    |    7 +-
>>>    randr/rrpointer.c                   |    2
>>>    randr/rrprovider.c                  |   25 +++++++
>>>    randr/rrscreen.c                    |    2
>>>    10 files changed, 194 insertions(+), 9 deletions(-)



More information about the xorg-devel mailing list