[waffle] [PATCH 02/29] gbm: don't fetch gbm_bo_* symbols from libgbm

Frank Henigman fjhenigman at google.com
Thu Jan 7 10:32:39 PST 2016


On Thu, Jan 7, 2016 at 11:20 AM, Emil Velikov <emil.l.velikov at gmail.com> wrote:
> On 7 January 2016 at 00:19, Frank Henigman <fjhenigman at google.com> wrote:
>> On Wed, Jan 6, 2016 at 4:27 PM, Emil Velikov <emil.l.velikov at gmail.com> wrote:
>>> Neither one of these is (or has been) used. Remove them for now and
>>> introduce back when needed.
>>>
>>> Presently this causes run-time issues whenever someone attempts to use an
>>> old version of libgbm (mesa circa 9.2) as the gbm_bo_get_fd symbol is
>>> missing.
>>>
>>> Due to the multiple gbm provides and their inconsistent versioning
>>> (xxx: check ?) we cannot reliably add a limitation during the configure
>>> stage.
>>>
>>> This commit reverts a hunk of commit 14e3356a01d(gbm: make platform
>>> friendlier to derived classes). The latter of which could have kept
>>> these as a separate patch :)
>>>
>>> Cc: Frank Henigman <fjhenigman at google.com>
>>> Cc: Chad Versace <chad.versace at intel.com>
>>> Signed-off-by: Emil Velikov <emil.l.velikov at gmail.com>
>>> ---
>>>  src/waffle/gbm/wgbm_platform.h | 11 +----------
>>>  1 file changed, 1 insertion(+), 10 deletions(-)
>>>
>>> diff --git a/src/waffle/gbm/wgbm_platform.h b/src/waffle/gbm/wgbm_platform.h
>>> index 1a08183..edcbe95 100644
>>> --- a/src/waffle/gbm/wgbm_platform.h
>>> +++ b/src/waffle/gbm/wgbm_platform.h
>>> @@ -41,16 +41,7 @@
>>>      f(struct gbm_surface *, gbm_surface_create           , (struct gbm_device *gbm, uint32_t width, uint32_t height, uint32_t format, uint32_t flags)) \
>>>      f(void                , gbm_surface_destroy          , (struct gbm_surface *surface)) \
>>>      f(struct gbm_bo *     , gbm_surface_lock_front_buffer, (struct gbm_surface *surface)) \
>>> -    f(void                , gbm_surface_release_buffer   , (struct gbm_surface *surface, struct gbm_bo *bo)) \
>>> -    f(struct gbm_bo *     , gbm_bo_create                , (struct gbm_device *gbm, uint32_t width, uint32_t height, uint32_t format, uint32_t flags)) \
>>> -    f(void                , gbm_bo_destroy               , (struct gbm_bo *bo)) \
>>> -    f(int                 , gbm_bo_get_fd                , (struct gbm_bo *bo)) \
>>> -    f(uint32_t            , gbm_bo_get_width             , (struct gbm_bo *bo)) \
>>> -    f(uint32_t            , gbm_bo_get_height            , (struct gbm_bo *bo)) \
>>> -    f(uint32_t            , gbm_bo_get_stride            , (struct gbm_bo *bo)) \
>>> -    f(uint32_t            , gbm_bo_get_format            , (struct gbm_bo *bo)) \
>>> -    f(union gbm_bo_handle , gbm_bo_get_handle            , (struct gbm_bo *bo)) \
>>> -    f(struct gbm_device * , gbm_bo_get_device            , (struct gbm_bo *bo))
>>
>> My surfaceless branch uses these.  The branch I've been thinking I'll
>> send to the list "soon" for ages.
>> I'll add in a check for gbm_bo_get_fd and disable surfaceless if not
>> found (though I'm sure it wouldn't work anyway on ancient mesa).
>> In the mean time if you have to remove this stuff go ahead, or maybe
>> just remove gbm_bo_get_fd. Then I won't have to add the others back
>> in.
> As you get the chance to re-spin the surfaceless patches, feel free to
> revert it straight away. I doubt there'll be many changes that
> conflicts in the near future so things should be safe. Until then
> let's nuke [everything] that we don't use. Please ?

Sure, I can add back what I need.  Or try to land surfaceless first.  (^:

Reviewed-by: Frank Henigman <fjhenigman at google.com>


More information about the waffle mailing list