[Intel-gfx] [PATCH i-g-t] core_prop_blob ioctl_wrappers: Fix new tests/benchmarks for android

Morton, Derek J derek.j.morton at intel.com
Wed Oct 14 09:42:00 PDT 2015

>On 13 October 2015 at 16:35, Daniel Vetter <daniel at ffwll.ch> wrote:
>>> +void igt_require_propblob(int fd)
>>> +{
>>> +     struct local_drm_mode_create_blob c;
>>> +     struct local_drm_mode_destroy_blob d;
>>> +     uint32_t blob_data;
>>> +     c.data = &blob_data;
>>> +     c.length = sizeof(blob_data);
>>> +
>>> +     igt_require(drmIoctl(fd, LOCAL_DRM_IOCTL_MODE_CREATEPROPBLOB, &c) == 0);
>>> +     d.blob_id = c.blob_id;
>>> +     igt_require(drmIoctl(fd, LOCAL_DRM_IOCTL_MODE_DESTROYPROPBLOB, 
>>> +&d) == 0); }
>> If you want to do this in the library the usual way is to wrap the 
>> ioctls as functions and put the relevant -ENOTTY check in there as an 
>> igt_require, followed by an igt_assert for anything else that might 
>> have gone wrong.
>> I'd just keep this in the test as a static function though, since then 
>> you don't have to write api docs ;-)
>I have no real opinion either way; apologies for breaking the build!
>Is Android up-to-date with libdrm, or? If it's not, I suspect kms_atomic.c should just be excluded from the Android build altogether.

The kms_* tests are dependant on cairo which is not in the android build by default. Personally I have never tried getting cairo to run on android so have never built the kms_* tests for android.

The version of libdrm used in android will match the kernel version being used in the particular android project. (So it would not be the latest version that goes with drm-nightly)

You aren't the 1st person to break the android build (and won't be the last either) :-)



More information about the Intel-gfx mailing list