[21362] in Athena Bugs
Packages split between the srvd and local disk cause autoconf problems
daemon@ATHENA.MIT.EDU (Mitchell E Berger)
Fri Jan 24 19:15:52 2003
Message-Id: <200301250015.TAA09470@defender.mit.edu>
To: bugs@MIT.EDU
Date: Fri, 24 Jan 2003 19:15:46 -0500
From: Mitchell E Berger <mitchb@MIT.EDU>
The other day, on an Athena 9.1.18 Sun, I was attempting to build a piece
of software that uses autoconf (it happened to be galeon, but that's
unimportant). One of the checks it made was to see whether the machine had
liboaf >= 0.6.5 (which we do).
configure dutifully complained that 'oaf-config --version' reported 0.6.10,
while the 0.6.8 library was actually found. Neverminding that either of those
would have been fine, the configure blew out.
The problem is that liboaf was upgraded from 0.6.8 in Athena 9.1.18 to 0.6.10
in 9.1.22 (or an in-between Athena version). The machine in question was
AUTOUPDATE=false, and so since /usr/athena/lib/liboaf.so.0.0.0 is a file on
local disk, it was still version 0.6.8. Meanwhile, oaf-config (and I believe
also the relevant header file, which configure would have checked if it got
past the above test) is a symlink into the srvd, and since Athena 9.1.22 had
been released, it reported 0.6.10.
This is a bug; all pieces of a package that depend on each other's being at
a particular version need to be together, regardless of whether they are both
on the srvd or both on localdisk. True, building galeon is not something
that's normal for a Solaris Athena user to be doing, but we do support being
able to build arbitrary autoconf software against the libraries we put in the
release, and we suggest that setting AUTOUPDATE=false will not cause versions
of release software to change on you.
Presumably, the specific oafd issue could be fixed by messing with sys_rvd.big,
but the generic issue is that finding all places where this type of problem
exists is sort of intractable. It will become a non-issue when we don't have
system packs anymore in 9.2, but in the meantime, does anyone think we
can/should do anything about it?
Mitch
