Re: (no subject)

On 06/10/2010 05:16 PM, Gregory Maxwell wrote:
> It's worth elaborating on this, because I don't want people thinking
> that libtheora has known and unfixed problems accessing uninitialized
> variables.
>
>
> In libtheora there are a number of places where an initialization
> isn't obvious to a compiler, usually due to cross function boundaries.
> Some of these places are performance critical.
>
> In order to suppress compiler warnings in places where they would
> incorrectly warn of missing initialization libtheora initializes a
> variable to itself. e.g.
>
> int a=a;
>
> This odd looking code causes GCC (for example) to emit nothing at all,
> so it's the same as "int a;" but it suppresses the warning.
>
> For some odd reason, the Microsoft tool-chain actually generates code
> for the a=a case and then the runtime checker warns on it.  It's a
> harmless warning, however.
>
> The valgrind tool we use on GNU/Linux is far more intelligent with
> uninitialized access checks and isn't confused by this at all.
>   
Greg,

Thank you for taking the time to explain this.   I think it's due to my
ignorance of C that I am not aware of such issues, so when I see a
plethora of warnings here-and-there in some cases (not necessarily with
libtheora, but perhaps other packages), I always wonder, "Why did the
devs do that?!", as I tend to be quite pedantic and would imagine that a
warning means "don't do that".

Apparently this is not always the case, and it's just the nature of C
that it's not always clear what is to be expected, at least to the compiler.