Re: [PATCH] Adjust testsuite and rope for "Deprecated implicit typename"

[Paolo Carlini <pcarlini at unitus dot it>]

Phil Edwards wrote:

> http://gcc.gnu.org/simtest-howto.html
>
> I usually use powerpc-eabisim, just because I think that chip is cool.  :-)
>
> Any platform which has a lot of trouble is also a good one to monitor;
> sparc-sun-solaris2* somewhat falls into this camp for me, since I have
> some of those where I work, and so I have a vague interest in keeping
> those functioning.  But for that one I can normally just try it natively.
> The *BSD platforms are also important one to keep functioning.

Ok. Thanks for your precious indications.

> > > > diff -urN libstdc++-v3-orig/include/ext/ropeimpl.h
> > >
> > > Also as a side note, the -p option to diff or cvs diff is usually
> > > appreciated.  For libstdc++, most of the time the "function" shown turns
> > > out to be something like "namespace std {" but every so often diff gets
> > > the right name.  :-)
> >
> > Another very good suggestion (much easier to put in practice ;-).
> > No one among the CVS developers is working on improving the support for C++ (or
> > other programming languages for that matter) ??
>
> Well, CVS just uses the GNU diff code.

Yes, of course. I was about to send an errata s/CVS developers/diff developers/

> The GNU diff code only looks for a matching regular expression higher
> up in the file than where the diff occurs.  I looked at this briefly;
> in the CVS source it's diff/context.c:find_function().
>
> None of that code is language-aware (other than "it looks like C source"),
> and it's not really possible to distinguish the cases based only on
> regexp matching.  It would have to use some kind of actual language parser
> or codewalker.
>
> I wonder how well -p works on Java.

I see. And (function) templates are not the easiest thing to parse... ;-)

Cheers,
Paolo.