Fix more sdktools crash on 64 bits

[bottiger1]

Found some more pointer arithmetic in sdktools assuming 32bits.

[KyleSanderson]

is there still - 4 to get IClient?

[bottiger1]

is there still - 4 to get IClient?

I did grep "+ 4" and "- 4" in the sdktools directory. I did not check any other directory.

[asherkin]

While I understand it seems to be the same in practice, sizeof(intptr_t) feels wrong here - I would expect sizeof(void *) for this usage.

[assyrianic]

While I understand it seems to be the same in practice, sizeof(intptr_t) feels wrong here - I would expect sizeof(void *) for this usage.

Any particular reason why? Most of the code itself is already using [u]intptr_t anyway.

[asherkin]

Any particular reason why? Most of the code itself is already using [u]intptr_t anyway.

Which is obviously correct for the math, but the spec only appears to guarantee sizeof(intptr_t) >= sizeof(void *) - and what we want here is the size of a (specific) pointer, not the size of the integer type to represent any arbitrary pointer.

[bottiger1]

We don't need to have a long discussion about this. If you want to change it to void* for a hypothetical architecture where the fastest integer type is somehow larger than the pointer type, it doesn't matter to me. I just want these crashes gone.

[dvander]

static_assert() that your intention is true and move on, I say.