This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Fix collate_members_char[wchar_t].cc


Benjamin Kosnik wrote:

> > > > > > -  VERIFY(i1 == i2);
> > > > > > +  VERIFY(i2 == 1);
> > >
> > > The original code is correct and according to what you just said it
> > > would work for you as well.  So dump the change and use the original
> > > VERIFY.
> >
> > Sorry. I think there is a misunderstanding. I cannot use the original VERIFY
> > because string::compare does /not/ normalize! (I don't have the standard here,
> > but, f.i., Josuttis, p. 511).
>
> So, shouldn't it be
>
>  +  VERIFY(i2 > 0);
>
> ?

Hi Benjamin. Well, I think VERIFY(i2 == +1) (or -1 in the other testcase) is
better, in that, whereas the value computed by string::compare is /not/ normalized,
that of collate::compare /is/ normalized (it wasn't in v3, wrongly, before mine and
Uli's recent patch)

Do you agree?

Ciao, Paolo.



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]