On 4/8/22 08:04, Giovanni Mascellani wrote:
diff --git a/libs/vkd3d-shader/hlsl.c b/libs/vkd3d-shader/hlsl.c index eabe189f..1ebd97c1 100644 --- a/libs/vkd3d-shader/hlsl.c +++ b/libs/vkd3d-shader/hlsl.c @@ -1188,7 +1188,7 @@ static void dump_ir_constant(struct vkd3d_string_buffer *buffer, const struct hl break;
default:
assert(0);
vkd3d_unreachable("Unknown type"); } } if (type->dimx != 1)
I don't think there's value in specifying a message for cases like this, and I'd rather not have to do so in new code. This applies to pretty much anywhere we have an assert() in a switch default case.
(Ideally we should avoid the 'default' keyword where possible, although in some cases this is difficult.)
@@ -1542,12 +1534,10 @@ static void write_sm4_cast(struct hlsl_ctx *ctx, break;
case HLSL_TYPE_BOOL:
/* Casts to bool should have already been lowered. */assert(0);break;
vkd3d_unreachable("Casts to bool should have already been lowered");
This one I think should remain an assert. I can more easily see this getting triggered due to some forgotten implicit agreement. Hence we probably shouldn't tell the compiler this code path is unreachable.
Using an assert(!"string") or assert(0 && "string") construction (or defining our own macro) seems reasonable to me, though.