Bugzilla Bug 108534

While emerging dev-db/qdbm-1.8.33, portage fails with the following;

making executable: /usr/lib/libjqdbm.so.1.0.0
making executable: /usr/lib/libqdbm.so.11.5.0

QA Notice: the following files contain insecure RUNPATH's
 Please file a bug about this at http://bugs.gentoo.org/
 For more information on this issue, kindly review:
 http://bugs.gentoo.org/81745
/var/tmp/portage/qdbm-1.8.33/work/qdbm-1.8.33/perl/depot/../..:/lib
usr/lib/perl5/site_perl/5.8.7/sparc-linux/auto/Depot/Depot.so
/var/tmp/portage/qdbm-1.8.33/work/qdbm-1.8.33/perl/curia/../..:/lib
usr/lib/perl5/site_perl/5.8.7/sparc-linux/auto/Curia/Curia.so
/var/tmp/portage/qdbm-1.8.33/work/qdbm-1.8.33/perl/villa/../..:/lib
usr/lib/perl5/site_perl/5.8.7/sparc-linux/auto/Villa/Villa.so


!!! ERROR: dev-db/qdbm-1.8.33 failed.
!!! Function dyn_install, Line 1058, Exitcode 0
!!! Insecure binaries detected
!!! If you need support, post the topmost build error, NOT this status message.

Created an attachment (id=70203) [edit]
makeMake is bad

This patch WORKSFORME

i havent looked, but cant you add that var to RUNENV rather than changing every
call to `make` ?

hattya, please bump with patch

Created an attachment (id=70293) [edit]
qdbm-runpath.patch

how about this instead

SpanKY: we'll probably have to commit that ourselves as hattya looks kinda MIA.

Fixed in CVS.

Still broke, same thing as originally posted.

Guess we are stuck here. Jason: the first version of the patch was fixing it
for
you ?

Correct, the original patch works for me on the systems where it has been
problema tic.  Some more notes on that;

I have three systems I've been testing this on (x86, amd64 and sparc64).  All
systems are running up to date ~arch systems which are updated on a daily basis
and all have the perl use flag enabled.  The ~x86 and ~sparc64 systems are
running 32 bit userlands where the ~amd64 system is running a 64 bit userland. 
All have pax-utils installed.

The ~amd64 system has not had a single problem with qdbm at all, before and
after the patch.  However both the ~x86 and ~sparc64 systems have had problems
with the unpatched qdbm and the patched version currently in portage.  The first
patch in this bug (cuurrently labeled makeMake is bad), works on those two
systems though.

vapier: We'll probably go with the "makeMake is bad" patch if you don't see why
yours failed.

hattya: could you apply the "makeMake is bad" patch instead ? Looks like it's
the only one that solves it.

in CVS.

Jason: can you confirm it's ok with the one in CVS ?
hattya: could you revbump the ebuild so that people affected pick up the change
in  normal upgrades ?

The ebuild currently in portage fails in the same fashion.  Also, it appears to
still be using the second patch.  

Yes, the patch in CVS looks unchanged.

Hattya, could you please doublecheck and commit the "makemake is bad" attached
patch ?

Sorry, I forgot to commit...
In CVS now.

Thx Hattya. 
 
Arches please test and mark stable. 

amd64 stable

x86 done

Stable on SPARC

Alpha stabled.

Marked ppc64 stable.

In testing with FEATURES="test" the tests did fail like so:

LD_LIBRARY_PATH=.:/lib:/usr/lib:/var/tmp/portage/homedir/lib:/usr/local/lib 
./cbtest misc
<Miscellaneous Test>

Checking serialization of list ... ok
Checking serialization of map ... ok
Checking string utilities ... ok
Checking encoding utilities ... ok
Checking date utilities ... ./cbtest: W3CDTF formatter is invalid
make: *** [check] Error 1

!!! ERROR: dev-db/qdbm-1.8.33-r2 failed.

Does this warrant any attention?  1.8.30 failed in a similar fashion.  I will
leave ppc64 on the cc and we can discuss or split off to a different bug.

Stable on ppc.

Apparently ppc64 stable keyword was lost somewhere along the way.

Remarked stable; I dont know what happened.  The above stated test failure
still
exists however.

Ready for GLSA.
About the test failure: probably better to open a separate bug about it, since
we'll close this one when the security part will be done.

GLSA 200511-02
ia64 should mark stable to benefit from GLSA