Re: Reason for linking the GNAT tools dynamically
On Sat, 2012-03-31 at 12:05 +0200, Ludovic Brenta wrote:
> Svante Signell <firstname.lastname@example.org> writes:
> > Any ideas of the reason for static/dynamic switch in patch:
> > ada-link-lib.diff? Without the reason for it being introduced, there is
> > not much incitament to submit patches to it...
> This is explained in /usr/share/doc/gnat-4.6/README.Debian:
> - Install the shared Ada libraries as '.so.1', not '.so' to conform
> to the Debian policy.
> - Don't include a runtime link path (-rpath), when linking binaries.
> - Instead of building libada as a target library only, build it as
> both a host and, if different, target library.
Now libada is not built at all, at least not by defauilt, why was it
> - Build the GNAT tools in their top-level directory; do not use
> recursive makefiles.
> - Link the GNAT tools dynamically.
> - Fix a bug in src/gnattools/configure.ac whereby a nonexistent version
> of indepsw's body was selected. Regenerate configure. (PR ada/27300)
> libgnatvsn and libgnatprj guarantee the compatibility of GNAT with other
> tools built from other source packages like asis and gnat-gps. Linking
> the GNAT tools against libgnat-4.6.so is a prerequisite for that and
> additionally saves about 10 MiB on an installation.
I know libgnatvsn and libgnatprj are needed for other tools to work.
Any reson why upstream hav not taken that approach, accepted the
patches? Of course linking dynamically is good to save space. But why
was the gnattools-cross removed by these patches??