ftbfs in artful due to DT_RPATH/DT_RUNPATH fun and games

[Michael Hudson-Doyle]

elektra fails to build in artful with tests failing like this:

Traceback (most recent call last):
  File "/<<PKGBUILDDIR>>/src/bindings/swig/python/tests/test_kdb.py", line 61, in tearDownClass
    with kdb.KDB() as db:
  File "/<<PKGBUILDDIR>>/obj-x86_64-linux-gnu/src/bindings/swig/python/kdb.py", line 1571, in __init__
    _kdb.KDB_swiginit(self, _kdb.new_KDB(*args))
kdb.KDBException: 1 Warning was issued:
 Warning number: 1
 Description: could not load module, dlopen failed
 Ingroup: modules
 Module: dl
 At: /<<PKGBUILDDIR>>/src/libloader/dl.c:80
 Reason: of module: libelektra-resolver.so, because: libelektra-resolver.so: cannot open shared object file: No such file or directory
 Mountpoint:
 Configfile:
Error (#40) occurred!
Description: Failed to open default backend (see warnings for more information)
Ingroup: kdb
Module:
At: /<<PKGBUILDDIR>>/src/libelektra/kdb.c:204
Reason: could not open default backend
Mountpoint:
Configfile:

After some digging, it turns out that this is because the _kdb.so python extension module has a DT_RUNPATH pointing at the directory containing libelektra-resolver.so, not DT_RPATH (which would work, see http://blog.qt.io/blog/2011/10/28/rpath-and-runpath/).

Nothing in the packaging afaict has an opinion on rpath vs runpath so this is probably down to some change in some other package.

That said, using -rpath to make tests pass when linking files that get installed to locations on the default so search path smells a bit. Maybe it would be better to set LD_LIBRARY_PATH when running the tests?

[Michael Hudson-Doyle (mwhudson)]

Oh sigh debian has already been through this: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=846200