On Tue, Apr 06, 2021 at 03:32:48PM +0300, Efraim Flashner wrote:
>  
> +       ;; The llvm-based tests are flakey on non-Intel hardware.
> +       #:tests? ,(if (string=? "powerpc-linux" (or (%current-system)
> +                                                   (%current-target-system)))
> +                   '#f '#t)
> +
> 

Looks like Charlène of OpenBSD (I think?) is right, there are some
endianness bugs in mesa with llvm. I've tried, building mesa for powerpc
without llvm isn't actually possible, so disabled tests for now.

Snippet from the test suite log:

Testing util_format_a1b5g5r5_unorm_unpack_rgba_8unorm ...
FAILED: {0x39, 0xc5, 0x00, 0x00} obtained
        {0x00, 0x00, 0xff, 0x00} expected
FAILED: {0xc5, 0x00, 0x18, 0xff} obtained
        {0x00, 0xff, 0x00, 0x00} expected
FAILED: {0x00, 0x18, 0xe6, 0x00} obtained
        {0xff, 0x00, 0x00, 0x00} expected
FAILED: {0x00, 0x20, 0x00, 0x00} obtained
        {0x00, 0x00, 0x00, 0xff} expected
Testing util_format_a1b5g5r5_unorm_norm_flags ...
Testing util_format_x1b5g5r5_unorm_fetch_rgba_float ...
FAILED: {0.225806, 0.774194, 0.000000, 1.000000} obtained
        {0.000000, 0.000000, 1.000000, 1.000000} expected
FAILED: {0.774194, 0.000000, 0.096774, 1.000000} obtained
        {0.000000, 1.000000, 0.000000, 1.000000} expected
FAILED: {0.000000, 0.096774, 0.903226, 1.000000} obtained
        {1.000000, 0.000000, 0.000000, 1.000000} expected
FAILED: {1.000000, 0.870968, 1.000000, 1.000000} obtained
        {1.000000, 1.000000, 1.000000, 1.000000} expected
Testing util_format_x1b5g5r5_unorm_pack_rgba_float ...
FAILED: 00 3e obtained
        3e 00 expected
FAILED: 07 c0 obtained
        c0 07 expected
FAILED: f8 00 obtained
        00 f8 expected
FAILED: ff fe obtained
        fe ff expected


-- 
Efraim Flashner   <efraim@flashner.co.il>   אפרים פלשנר
GPG key = A28B F40C 3E55 1372 662D  14F7 41AA E7DC CA3D 8351
Confidentiality cannot be guaranteed on emails sent or received unencrypted