[PATCH 0/1] MR247: tests: Cast abs() result to unsigned int in compare_float().

[PATCH 1/1] tests: Cast abs() result to unsigned int in compare_float().

[Zebediah Figura]

From: Zebediah Figura

---
 tests/utils.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tests/utils.h b/tests/utils.h
index 5fd7b086e..a74275e3a 100644
--- a/tests/utils.h
+++ b/tests/utils.h
@@ -101,7 +101,8 @@ static bool compare_float(float f, float g, unsigned int ulps)
     if (y < 0)
         y = INT_MIN - y;

-    if (abs(x - y) > ulps)
+    /* Cast to unsigned int in case x - y == INT_MIN. */
+    if ((unsigned int)abs(x - y) > ulps)
         return false;

return true;

[Zebediah Figura (＠zfigura)]

I'm not sure if this is the best solution to this, maybe there's something cleverer, but this at least stops the function from treating -1.0 and 4.0 as equal.

[Chip Davis (＠cdavis5e)]

On Sun Jun 25 06:07:44 2023 +0000, Zebediah Figura wrote:
> I'm not sure if this is the best solution to this, maybe there's
> something cleverer, but this at least stops the function from treating
> -1.0 and 4.0 as equal.

I thought I wrote an `absdiff()` function for this exact scenario...

Oh yeah, it's in the d3d9 tests. It's not terribly clever. All it does is make sure the greater of the two operands is the minuend:

```c
static inline unsigned int absdiff(unsigned int x, unsigned int y)
{
    return x > y ? x - y : y - x;
}
```

~~See `wine/dlls/d3d9/tests/visual.c:66`.~~ EDIT: Wait, that one didn't go in yet. I could've sworn I had one like that in...

[PATCH 0/1] MR247: tests: Cast abs() result to unsigned int in compare_float(). - approved

[Giovanni Mascellani (＠giomasce)]

This merge request was approved by Giovanni Mascellani.

[Henri Verbeet (＠hverbeet)]

> I thought I wrote an `absdiff()` function for this exact scenario...
>
> Oh yeah, it's in the d3d9 tests. It's not terribly clever. All it does is make sure the greater of the two operands is the minuend:
> static inline unsigned int absdiff(unsigned int x, unsigned int y)
> {
>     return x > y ? x - y : y - x;
> }
>
> ~~See~~ `wine/dlls/d3d9/tests/visual.c:66`. EDIT: Wait, that one didn't go in yet. I could've sworn I had one like that in...

Well, we have compare_uint() in the various Wine D3D tests. So does vkd3d, and we could probably use "return compare_uint(x, y, ulps);" here.