3DC format test failures

[Stefan Dösinger]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Matteo,

I'm getting test failures in test_3dc_formats on various machines. Do
you have any idea how to change the test without reducing its
usefulness? It seems that we're seeing all possible swizzles on Windows.

Geforce 9600, Windows 7:
visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
got 0x007f7f7f, case 0.
visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
got 0x007f7f7f, case 1.

Radeon X1600, Windows 7:
visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
got 0x007fffff, case 0.
visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
got 0x007fffff, case 1.

This one smells like an implementation problem: Wine or r300g. Looks
like the draw fails.
visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
got 0x00ff00ff, case 0.
visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
got 0x00ff00ff, case 1.

Cheers,
Stefan
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
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=ZgiK
-----END PGP SIGNATURE-----

[Matteo Bruni]

2015-02-12 13:06 GMT+01:00 Stefan Dösinger <stefandoesinger(a)gmail.com>:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hi Matteo,
>
> I'm getting test failures in test_3dc_formats on various machines. Do
> you have any idea how to change the test without reducing its
> usefulness? It seems that we're seeing all possible swizzles on Windows.
>
> Geforce 9600, Windows 7:
> visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
> got 0x007f7f7f, case 0.
> visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
> got 0x007f7f7f, case 1.
>
> Radeon X1600, Windows 7:
> visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
> got 0x007fffff, case 0.
> visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
> got 0x007fffff, case 1.

Yes, I remember you mentioning that. I wrote a couple patches to
accept those two results too, but I was thinking about ignoring the G
and B components for ATI1N altogether (and do something about the
alpha part) as an alternative.
So essentially those patches got stuck in my private tree. Do you have
a preference on how to fix the test?
> This one smells like an implementation problem: Wine or r300g. Looks
> like the draw fails.
> visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
> got 0x00ff00ff, case 0.
> visual.c:16804: Test failed: Expected color 0x003f3f3f or 0x007f0000,
> got 0x00ff00ff, case 1.

Yes. Maybe ATI1N isn't supported by the HW? I don't remember which
exact r300-class GPU you have but it looks like ATI1N is only
supported on R500+, although AFAICS Mesa correctly doesn't report
ARB_texture_compression_rgtc with older GPUs. Can you check that the
extension (and the EXT_ variant) is or isn't reported as it would be
expected?

Cheers,
Matteo.
> Cheers,
> Stefan
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v2
>
> iQIcBAEBAgAGBQJU3JdCAAoJEN0/YqbEcdMwK1EP/RqbMOXv8wkirSvrudJuzOdk
> /R68iLVBN3wJSuTnGz+gJz4RR3N3hdN3YeI3M+eTHoyqk/4JhKDKSE8M/W94ILEh
> BYH1m98D/EpphSwkiREcRma4eAFkC3K4U0zvxjOLkLq0BLaK6vAD6BuK97jORqmh
> i8oZS4vIpayq5DSvi+niNz5Wk/JLo8fu/6049kmY0r7asSYOcl8LEAgynUKRiKDT
> 6wyEsnC2Y5lKJ4zRMHGmxi/68R3CsxgAODIIW/JKC6/XskEydAM5TZ+h4lOLj0zU
> fzAkJRwBpwP4nS3XEI2bbMyA5ZLCLfa7Qo+otuUwU6vL6+6v6uwBRWNJtW+rQydk
> 2ceSfEj+FaoUqlqI8iySRmVkivYDKyp4wZL8/2WlAqstvC40cbYKRsa9YtMZFZF8
> WpJip57t2pjAHdHCnAaA5qD7ulrjleAnMq7Yw3WqusA7NWPGWB8W02oKWuzfXYK3
> C2byraxi5HQaHfxi5c60+WdJyzKGUhDn9PBdDRHmWJF77msk9x4ktl/DvtBF9XNZ
> X549ff8otGpJvczCDJZObvLU4qvQmPw6raHda+8qtGj4HcqVu3rq+WNv7+7zvmxk
> myLKNhYNhxSYzwwLHOTO3XGac2VJ/Gl4BEiBCn1chIAuZRGd2iWB4DMD0ptq4ovw
> 0CiBK8gTt/f5TsLTTzTl
> =ZgiK
> -----END PGP SIGNATURE-----

[Stefan Dösinger]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Am 2015-02-12 um 16:59 schrieb Matteo Bruni:
> So essentially those patches got stuck in my private tree. Do you 
> have a preference on how to fix the test?
> No particular preference.
> Yes. Maybe ATI1N isn't supported by the HW?
> No, it works. It's the same Radeon X1600 where the Windows result is
> coming from.

I have one test that creates an invalid command stream that's rejected
by the kernel. It may be this one, I have to check.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
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=FprE
-----END PGP SIGNATURE-----