Re: GNAT 3.15p

[Ludovic Brenta <ludovic.brenta@insalien.org>]

Florian Weimer <fw@deneb.enyo.de> writes:

> Ludovic Brenta <ludovic.brenta@insalien.org> writes:
> 
> > The real problem is the soname.  I'm willing to change the soname to
> > "libgnat-3.15p.so" (removing the .1) to become more compatible with
> > ACT's GNAT, but does that violate the Debian Policy?
> 
> The users will receive annoying warnings from ldconfig, so this is not
> an option.

I tried it on the binary distribution of GNAT 3.15p.  I did not get
any warnings:

$ mkdir -p ~/tmp/usr/lib ~/tmp/etc ~/tmp/lib
$ touch ~/tmp/etc/ld.so.conf
$ cp /opt/gnat-3.15p/lib/gcc-lib/i686-pc-linux-gnu/2.8.1/adalib/lib* ~/tmp/usr/lib
$ /sbin/ldconfig -r ~/tmp -v
/lib:
/usr/lib:
        libgnat-3.15.so -> libgnat-3.15.so
        libgnarl-3.15.so -> libgnarl-3.15.so

All ldconfig does is create a symlink from the soname to the file name
of the library.  Here, they happen to be the same, so no symlink is
even created.  What kind of warnings were you referring to?  Perhaps
your experience differs from mine?

(Yes, I also tried it on the libraries I compiled, with the .1 in
their soname. I did not get any warnings there either).

> > If and when gnat-3.2 and -3.3 support shared libraries,
> 
> When properly compiled, upstream should support shared libraries
> 
> > they should do so with a proper soname such as "libgnat.so.3.3".
> > Florian, as a GCC maintainer, would you be willing to try and
> > persuade the GCC steering committee to this?
> 
> This is not a Steering Committee issue.  You should contact ACT
> directly.

Yes, ultimately this would be the Right Thing for future versions.  In
fact, I am quite confident that ACT will change the soname;
historically they've changed it with every release of their compiler.
They just don't appear consider the FSF version to be a "release" yet.

Meanwhile, gnat 3.15p has been released and will not change anymore,
so I still think my earlier proposal makes the most sense.

-- 
Ludovic Brenta.