Re: [Libreoffice] [PATCH] Simple fix for fontconfig.cxx to review (from cppcheck report)

Le Sun, 15 Jan 2012 06:48:28 +0100, Ivan Timofeev <timofeev.i.s@gmail.com>  
a écrit:

> Hi Mat,
>
> > > cppcheck reported this :
> > > core/vcl/generic/fontmanager/fontconfig.cxx
> > > 196 duplicateExpression style Same expression on both sides of '-'
> > >
> > > class SortFont : public ::std::binary_function< const FcPattern*, const
> > > FcPattern*, bool >
> > > {
> > > public:
> > > bool operator()(const FcPattern *a, const FcPattern *b)
> > > {
> > > int comp = compareFontNames(a, b);
> > > if (comp != 0)
> > > return comp < 0;
> > >
> > > int nVersionA=0, nVersionB=0;
> > >
> > > bool bHaveA = FcPatternGetInteger(a, FC_FONTVERSION, 0,
> > > &nVersionA) == FcResultMatch;
> > > bool bHaveB = FcPatternGetInteger(b, FC_FONTVERSION, 0,
> > > &nVersionB) == FcResultMatch;
> > >
> > > if (bHaveA && bHaveB)
> > > return nVersionA > nVersionB;
> > >
> > > 196: return bHaveA - bHaveA;<--- Same expression on both sides of
> > > '-'.
> > > }
> > > };
> > >
> > > it seems it should be :
> > > return bHaveA - bHaveB;
> > >
> > > Any idea ?
> > Subtracting booleans ??? Is there some magic here ?
> > Moreover, if we state that boolean difference returns a boolean, is
> > there an automatic cast, as the function should return an int ?
>
> AFAIK in arithmetic bools are concerted to int so that true->1 and  
> false->0.
> Then, the result is converted back to bool (because the function returns  
> bool) so that 0->false and 'any other value'->true.
>
> So "return bool1 - bool2" can be replaced with not-so-magic "return  
> bool1 != bool2", right?

For operator(), absolutely. I was actually talking about compareFontNames.  
But knowing that C++ does "on-the-fly" cast explains all.

> By the way, the author of the code is Caolan, let's ask him. :)
>
> Ivan
Mat