Run Microsoft Analyzer over dmd source code

[Walter Bright]

On 8/7/2011 10:11 AM, bearophile wrote:
> It contains FP warnings like the one I have asked for D too. DMD doesn't
> perform certain unsafe FP operations because they and break IEEE conformance,
> but casting a double to float is accepted and regarded "safe" (I am not sure
> of this): lexer.c(2500): warning C4244: 'initializing' : conversion from
> 'double' to 'float', possible loss of data


You mean implicit casting of double to float. Yes, it is accepted without 
complaint by dmd. The problem with requiring a cast is a cast is a blunt instrument:

    float f;
    double d;
    ...
    f = (float) d;

Now, suppose the maintenance guy decides to upgrade f to being a double to get 
more precision:

    double f;
    double d;
    ...
    f = (float) d;

and there's that cast to float he overlooked, sabotaging his upgrade. Even 
worse, suppose the type of f or d or both is changed to some user defined type, 
like BigFloat? That cast is just going to *cause* a bug, not fix it.

Requiring the programmer to load up on casts is not necessarily making the code 
less "bug-prone".