Re: Thanks!

[Paolo Carlini <pcarlini at unitus dot it>]

Benjamin Kosnik wrote:

> > ... therefore, I think Ulrich is suggesting that, whenever libc returns NUL, we
> > should simply fill grouping() with the empty string literal (""), right?
> >
> > In this way, everything should be ok.
>
> Thanks Ulrich.
>
> This sounds right. If you could work up a patch/testsuite entry I'd
> appreciate it as I'm in the middle of something else right now.

Sure. I'm already bootstrapping a minimal fix for the problem at hand.
Could you perhaps give me some hint on the most suited kind of testcase for this?
Right now, each locale which has _M_thousands_sep == '\0' leads to seg faults,
sooner or later... nothing very easy to test...

> I think I
> went through all the numpunct/moneypunct parts that use grouping info and
> checked for NULL grouping strings, but you might want to double check.

Sure.

> I think there is a little entry on the main locale docs page about this:
>
> http://gcc.gnu.org/onlinedocs/libstdc++/22_locale/howto.html
> http://gcc.gnu.org/onlinedocs/libstdc++/22_locale/locale.html
>
> See 6, Unresolved issues.
>
> Perhaps that could be updated as well?

Ok!

Cheers,
Paolo.