Re: Call for vetoes: Setting the global C++ locale in Gtk::Main

[Daniel Elstner <daniel kitta googlemail com>]

Hi,

Am Dienstag, den 06.04.2010, 13:01 +0400 schrieb Alexander Shaduri:

> Just checked it on FreeBSD 6.something, it is indeed an instance
> of std::runtime_error. Your patch should work fine.

Cool, thanks.

> I'm a bit worried about the
> #ifdef GLIBMM_EXCEPTIONS_ENABLED
> part though. Basically, it means that the no-exceptions build
> will fail on FreeBSD, right? I remember reading here that the
> whole no-exceptions stuff is going to be removed from *mm, so not
> sure if that is a problem.

It will definitely fail if libstdc++ has exceptions enabled but glibmm
doesn't.  It may also fail if libstdc++ has exceptions disabled -- no
idea what it does if it cannot throw the exception.

With the no-exceptions build, you get what you asked for.  There are
other instances where the no-exceptions build misbehaves, and the API is
different, too.  It only ever existed to serve the special needs of some
custom-tailored embedded platforms.

Given that there are ways to hack around the issue, e.g. by specifying
the C locale in the environment or by passing set_locale=false to the
Gtk::Main constructor, I see no need to compromise our proper API for
the benefit of the no-exception build.  And yes, it will indeed go away,
thanks heavens.

Actually, what I had in mind when I called for vetoes were reports of
incompatibilities with existing application code.  I did not anticipate
a discussion of badly broken target platforms. :-)

--Daniel