On Fri, Dec 07, 2007 at 08:14:38PM -0700, Martin Michlmayr <[EMAIL PROTECTED]>
was heard to say:
> * Daniel Burrows <[EMAIL PROTECTED]> [2007-10-16 19:56]:
> > > http://people.debian.org/~lucas/logs/2007/09/10/gcc43/libsigc++-2.0_2.0.17-2build1_gutsylp.buildlog
> >
> > http://www.nabble.com/Re:-libsigc++-and-g++-4.3-p11507869.html
>
> Is there any update on this, Daniel and Murray? I'm trying to build
> the Debian archive with 4.3 to report bugs and quite a few packages
> fail because of sigc++-2.0. Getting sigc++-2.0 fixed would make it
> much easier for me to see which packages have 4.3 related build
> issues.
>
> BTW, Fedora and OpenSUSE are quickly moving to gcc 4.3 so this needs
> to be fixed anyway.
Thanks for the ping. I haven't heard anything from the sigc++ folks,
but if we need to do something for Debian, it might be safe to remove
that typedef. I'm not sure because TBH, I'm not sure what the effect of
that typedef actually is or what it's trying to do! It looks like the
code is typedef'ing a name that's already used by a real class; I don't
know without checking whether the existing class or the new typedef wins.
I don't think I'll have time to look into this tomorrow. If someone
else does have time, I think the things to investigate are:
* does removing the typedef break code when compiling with 4.2?
* does removing the typedef break code linked against old versions?
* does removing the typedef and compiling with 4.2 allow code compiled
with 4.3 to link against the library?
* does removing the typedef and compiling with 4.2 allow code linked
against the modified library to link against a standard library?
* do the exported symbols and the symbols required by client code
remain stable when the typedef is removed?
The best case would be that the ABI is unchanged by this and all code
compiles without that typedef, in which case we can strip it from the
Debian package without any trouble. If not, well, things will be more
complicated.
Daniel
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
