Re: .: 2 commits - configure.in sw/source

On 15/02/12 19:28, Tor Lillqvist wrote:
> > I completely
> > fail to see what actual problem could be caused by this.
>
> And how is that relevant? By the same logic, you could write code that
> just "knows" that ints are always 32 bits and whatnot.

but no applicable standard requires ints to be 32 bits, so that is
nowhere near "the same logic".

> MSVC mangles struct and class differently (just one char difference,
> see http://en.wikipedia.org/wiki/Microsoft_Visual_C%2B%2B_Name_Mangling#Data_Type
> ) . But in the cases where these warnings are currently displayed that
> apparently doesn't happen, because if it would, it would have lead to
> linking errors obviously, and they would have been fixed earlier.

interesting, was unaware of that.

> Please, don't bother replying with potential references to the chapter
> and verse of some standard that might say that struct vs. classs
> should not cause differences in mangling of the type name. We use
> (more or less perfect) implementations of standards, not standards
> themselves.

indeed, given that "struct" vs "class" causes MSVC to actually do
something different (whether that difference is allowed by the standard
or not doesn't matter in practice), the warning makes sense.

> > could this warning be turned off? (it is MSVC, right?)
>
> Probably, but then, we could turn off all warnings one by one instead
> of fixing them... maybe we should start doing that. Whatever, I hate
> this crap.
>
> --tml