Re: lilypond-2.17.28 doesn't build with libc++

[Thomas Klausner]

On Wed, Nov 13, 2013 at 09:20:40AM +0100, David Kastrup wrote:
> Thomas Klausner <address@hidden> writes:
> 
> > Hi James!
> >
> > On Sat, Nov 09, 2013 at 09:39:34PM +0100, Joerg Sonnenberger wrote:
> >> On Sat, Nov 09, 2013 at 12:20:36PM +0000, James wrote:
> >> > These patches don't apply cleanly to current master.
> >> > 
> >> > If you can git format them, then I can shepherd them through our
> >> > patch review/test process for you.
> >> 
> >> Not easily and I don't have time for adopting them to a git checkout.
> >> Maybe Thomas has...
> >
> > Thanks for the offer!
> >
> > I've adapted the patches to apply to lilypond git (from yesterday).
> > There are two of them:
> 
> I'm not sure you've seen the issue in the tracker:
> <URL:http://code.google.com/p/lilypond/issues/detail?id=3656>.  It
> should cover the std-vector.hh part.

No, I hadn't seen that. The patch included there looks good, thanks.

> > The other patch removes ambiguity with respect to to_string.
> 
> Can you think of a reasonably nice way to have this ambiguity not occur
> in that manner?  The way your patch resolves the problem at the call
> points may be the simplest way to resolve it, but would likely lead to
> having the problem reintroduced inadvertantly repeatedly by people
> working with g++ and submitting new patches.

Good idea. But sorry, I don't speak C++ that well. Perhaps Jörg has an
idea?

> For a rarely used platform like clang++, the net result of this fix
> would likely be that we get to hear something like "oh, by the way, the
> last three versions did not compile using clang++ so I used GCC instead"
> occasionally.  Or not hear it at all.

Well, I'm currently regularly compiling lilypond with clang, and then
there's this rarely-used operating system called 'Mac OS X' that has
clang as its default compiler (and libc++ as its default C++ library
since Mavericks), so there might be a few more people who can report
issues :)
 Thomas