[gpsd-dev] Debian needs a release. I'm working on it.

[Eric S. Raymond]

Bernd requested a release for Debian by the beginning of September. I think
this is possible; I'm working on it.  I've already cleared most of the
urgent stuff from my mailbox.  As usual, the bug tracker looks pretty clean.

I took a long vacation from gpsd because (a) I could - the code is pretty
stable now, kept so by our regression-tests suite, and (b) all the arguments
about RPATH made my head hurt. Seriously.

Looking back at the RPATH thread in my mailbox, it still makes my head hurt.
I can't maintain what I don't really understand, and it is pretty clear that
trying to get clever with RPATH querying and setting wanders unto a swamp
where everyone has opinions and nobody has solutions.  Or at least not
solutions I comprehend - and I note the absence of any patches coming in
to address the problems.

So I have simplified the RPATH handling as much as possible.  Here
is what the build currently does:

1. If libdir (defaulting to /ust/local/lib) is not in {/lib,
   /usr/lib}, it is prepended to RPATH in the scons build environment.
   The attempt to probe the system path with ldconfig is gone (I heard
   the argument that that it screwed up cross-compilation).

2. In the SConstruct build environment, the build directory is
   prepended to RPATH only if chrpath=yes, which is *no longer the
   default*.  In this case chrpath edits the build directory out of
   RPATH in binaries before installing them in rootspace.

3. In the regress-driver script RPATH is always set to include the
   current directory. This means regressions can always be run regardless
   of the value of libdir or whether chrpath is on or off, but without
   affecting the load path compiled into binaries.

If anyone has a problem with these rules, explain.  Keep it simple.  Don't
make my head hurt.  And please change the mail subject to "RPATH".

I'm going to delete that thread from my mailbox now.  We start fresh.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>