[Mesa-dev] [PATCH v3 00/12] BSD portability (Meson, ANV, RADV, VC4/5, SWR)

[Greg V]

On 01/25/2018 15:07, Emil Velikov wrote:
> On 24 January 2018 at 18:02, Greg V <greg at unrelenting.technology> wrote:
>> meson: updated with suggestions from review.
>>
>> anon_file: added a debug_name argument for anv, removed posix_fallocate
>> and retry loops that were copy-pasted from weston, etc.
>>
>> SWR: since there's an environment variable KNOB_MAX_WORKER_THREADS that
>> lets it work without topology, it's fine to build on FreeBSD/DragonFly
>> as is (with this patch only).
>>
>> ETIME discussion on FreeBSD bugzilla:
>> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225324
>>
> Report looks pretty nice. I'm keeping my fingers crossed it goes well.
>
>> Greg V (12):
>>    meson: fix missing dependencies
>>    meson: fix BSD build
>>    meson: fix getting cflags from pkg-config
>>    meson: handle LLVM 'x.x.xgit-revision' versions
> Dylan pushed these.
>
>>    meson: glsl: add missing nir header dependency
> I think he's got an open question if we still need this one.
>
>>    util: add phys_memory.h for all memory measurements
>>    util: add anon_file.h for all memfd/temp file usage
>>    anv: remove unused Linux-specific include
>>    anv: add MAP_POPULATE fallback define for portability
>>    swr: support FreeBSD/DragonFlyBSD
> These are
> Reviewed-by: Emil Velikov <emil.velikov at collabora.com>
>
>>    util: emulate futex on FreeBSD using umtx
> Acked-by: Emil Velikov <emil.velikov at collabora.com>
Yeah, I'll push an update to this one — it's been discovered that we 
need to set a flag to interpret timeout correctly
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225415#c4
>>    clover: fix build with LLVM 5
>>
> There's an open question here.
Sure, I actually should've opened a bug report (about the build failure) 
instead of sending a patch.
> I'll give it another few day for people to skim through the
> util/anv/swr ones. Modulo any objections/explosions will push them to
> master + stable branches.
Cool. I think anon_file and phys_memory might need a bit of testing, but 
the anv/swr changes are simple preprocessor fixes that can indeed go to 
stable quickly.
> Thanks again for the great work Greg!
> Emil
Thanks :)