[110973] trunk/base/src/macports1.0/macports.tcl

[Clemens Lang]

On Mon, Sep 16, 2013 at 07:05:27AM -0700, Jeremy Huddleston Sequoia wrote:
> > This introduces a regression for the textmate2 port, which does link
> > against libc++, even when the C++ settings haven't been touched.
> > Please revert this and/or achieve what you want in a different way.
> 
> It looks like this is actually behaving correctly then.

It is behaving correctly in the sense that it detects textmate2 wasn't
linked against libstdc++. However, textmate2 doesn't even support
linking against libstdc++ and it did work just fine before – that makes
it a regression.

> Perhaps we might also want to add an override to skip that check which
> certain ports (like textmate2) can set...

We should probably rather add a property that holds the C++ runtime
being used by a port. However, none of this is possible within
rev-upgrade, because rev-upgrade doesn't have access to any information
from the Portfiles.

This check should really go into the post-destroot check phase that was
also written in GSoC 11 but not yet merged into trunk. In this phase,
the necessary context for overriding checks from a Portfile is available
and it makes much more sense to display a warning that the wrong C++
runtime was used after the build rather than after all builds.

-- 
Clemens Lang