Re: Compiling lilypond

[Mojca Miklavec]

Dear David,

On 17 July 2016 at 00:47, David Kastrup wrote:
>
> [For the record: compiling LilyPond with Clang++ currently fails]
>
> Mojca Miklavec writes:
>
>> Hi,
>>
>> It seems that this is also a problem on the latest OS X, not just on
>> my ancient box. We'll probably drop an email on the bug mailing list.
>
> Here is an example you can report to the Clang++ mailing list.  This
> overloading resolution seems to be what essentially fails with the
> current LilyPond source.

Thank you very much for the very clear minimal example.

I'll try to get this resolved (I'm still waiting for some initial
feedback from the MacPorts mailing list).

> Unfortunately, I don't see much of a way to circumvent this one in the
> current code base without completely throwing large reorganizations and
> simplifications away.
>
> Sorry, you'll need to use GCC for now.

Using GCC could be a serious problem (depending on whether or not
lilypond uses any C++ interface from its dependencies), mainly because
lilypond would then use libstdc++, while all other libraries use
libc++. If lilypond uses any C++ library call, this would be a perfect
recipe for crashes. But we need to test first anyway. Maybe it just
works.

(Please note that this is not a problem for the binary you distribute
because you *do* compile all dependencies with the same compiler, but
it is a problem for a package manager where we cannot provide five
copies of each library, with the only difference being the compiler
used.)

In an ideal world gcc would work with libc++. But if the problem gets
fixed in clang, it's no problem (for MacPorts) to depend on the latest
(even if development) version of clang.

Thank you,
    Mojca