[PATCH 1/2] wined3d: Correctly parse SM4 dcl_immediateConstantBuffer opcode.

[Henri Verbeet]

On 3 December 2015 at 11:04, Józef Kucia <jkucia at codeweavers.com> wrote:
> Signed-off-by: Józef Kucia <jkucia at codeweavers.com>
> ---
> This fixes infinite loop when dcl_immediateConstantBuffer is parsed.
>
> Unfortunately, I had to reindent whole wined3d_sm4_opcode enum in order to make
> space for WINED3D_SM4_OP_DCL_IMMEDIATE_CONSTANT_BUFFER. I was also considering
> to use a shorter name: WINED3D_SM4_OP_DCL_IMMEDIATE_BUFFER.
> ---
It's fine, if perhaps unfortunate. You could do something like
WINED3D_SM4_OP_DCL_IMM_CB if we really want something shorter, but
it's probably best to just follow the opcode name.

> +    WINED3D_SM4_OP_MAX                           = 0x34,
> +    WINED3D_SM4_OP_DCL_IMMEDIATE_CONSTANT_BUFFER = 0x35,
> +    WINED3D_SM4_OP_MOV                           = 0x36,
This is odd. I'd expect a DCL_ opcode to be in the > 0x58 range.
Still, there's a hole there in the normal range, and "min" and "max"
are also reversed compared to what you'd expect, so I suppose it's at
least plausible.

>      opcode_token = *(*ptr)++;
>      opcode = opcode_token & WINED3D_SM4_OPCODE_MASK;
> -    len = ((opcode_token & WINED3D_SM4_INSTRUCTION_LENGTH_MASK) >> WINED3D_SM4_INSTRUCTION_LENGTH_SHIFT) - 1;
> +
> +    if (opcode == WINED3D_SM4_OP_DCL_IMMEDIATE_CONSTANT_BUFFER)
> +        len = *(*ptr) - 1;
> +    else
> +        len = ((opcode_token & WINED3D_SM4_INSTRUCTION_LENGTH_MASK) >> WINED3D_SM4_INSTRUCTION_LENGTH_SHIFT) - 1;
>
I don't think this is the right place to handle this. You should be
able to handle this the same way as the other DCL_ opcodes. It should
probably also have an entry in the opcode_info table. (Compare e.g.
6d948e1a8cdd8a9a5ae2b6124468275cbf13f2b8.)