This changes one of the values of the integer clear color to
0x7f817f81. If this were interpreted as a float value then it would be
a signalling NaN. If this passes through an x87 register then the
store instruction will convert it to a quiet NaN by setting bit 22.
The integer value would then be incorrect.
This currently exposes a bug in the blorp code path in the i965 driver
in Mesa.
---
tests/spec/ext_texture_integer/fbo-integer-precision-clear.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/tests/spec/ext_texture_integer/fbo-integer-precision-clear.c
b/tests/spec/ext_texture_integer/fbo-integer-precision-clear.c
index f4bf452..eea70f6 100644
--- a/tests/spec/ext_texture_integer/fbo-integer-precision-clear.c
+++ b/tests/spec/ext_texture_integer/fbo-integer-precision-clear.c
@@ -129,7 +129,12 @@ test_fbo(const struct format_info *info)
GLenum status;
GLboolean intMode;
GLint buf;
- static const GLint clr[4] = { 300000005, 7, 6, 5 };
+ /* The last value is chosen because if it were to be
+ * interpreted as a float value then it would be a signalling
+ * NaN value. If this passes through an x87 floating-point
+ * register then it will be corrupted by the fst
+ * instruction. */
+ static const GLint clr[4] = { 300000005, 7, 6, 0x7f817f81 };
GLint pix[4], i;
bool pass = true;
--
1.9.3
_______________________________________________
Piglit mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/piglit
