swscale: Disallow conversion to GBRP16
authorMartin Storsjö <martin@martin.st>
Mon, 11 Feb 2013 11:37:29 +0000 (13:37 +0200)
committerMartin Storsjö <martin@martin.st>
Mon, 11 Feb 2013 18:03:16 +0000 (20:03 +0200)
commitd192ac3e03d1ebdb671f22b290da577d19f3e156
tree3a7eb8200a818953161eb43914dc20c5acc1272e
parent62a43d30cc3b9b2460f2c926beb522df32493625
swscale: Disallow conversion to GBRP16

This reverts parts of d6d5ef5534d582, that didn't work right. (The
tests that were added failed on big endian, and the output looked
garbled on little endian as well.)

This is due to the fact that the intermediate scaling values (from
e.g. hScale8To19_c or hScale16To19_c) are stored as int32_t and
thus requires a separate output function, while yuv2gbrp_full_X_c
only interprets it as int16_t.

Signed-off-by: Martin Storsjö <martin@martin.st>
libswscale/utils.c
tests/ref/lavfi/pixdesc
tests/ref/lavfi/pixfmts_copy
tests/ref/lavfi/pixfmts_null
tests/ref/lavfi/pixfmts_scale
tests/ref/lavfi/pixfmts_vflip