The config.py script doesn't seem to cope very well with symlinks, so
let's pass it the right derivation outputs wherever possible and fall
back to drv.out.
I've disabled the tests because they somehow now seem to cause the build
to fail even though the tests failed *before* the merge of the
closure-size branch, but the whole build didn't fail regardless.
Here is a build from before the closure-size branch merge:
http://hydra.nixos.org/build/34367296
If you have a look at the build log, you already see a bunch of failing
tests (to be exact: the same set of tests that are failing now with the
fix of the preConfigure phase).
Other than that, the build now succeeds on my machine.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Commit 0055c6a introduced a new preConfigure hook that sets the right
qmake path. Unfortunately the mkDerivation attributes of pyqt5 override
the whole configurePhase, so this hook isn't run at all.
This fixes the build of pyqt5 and it now successfully compiles on my
machine.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
I think what's happening is that the linker automatically adds DT_NEEDED dependencies to some libraries because it finds these libraries are being used directly, but
because they're not linked explicitly with -lflags, the gcc wrapper does not add them to RUNPATH.
Rewrite dlopening stuff in hacky way (due ctypes.util totally brokennes:
it attempt to use /sbin/ldconfig, gcc from PATH and other tricks to
detect sonames, I replaced it with simple table lookup)
Also I add patch to bypass another rounding regression in tests
(this patch submitted upstream as well)
Fixes#12663: problems in python stuff due to old timestamps in sources.
- Files in sources older than a certain year are set to that year.
- Applied with 1980 for all python packages due to the way it often uses zip.
(cherry picked from staging commit e4ab8aee62)
Fixes#12663: problems in python stuff due to old timestamps in sources.
- Files in sources older than a certain year are set to that year.
- Applied with 1980 for all python packages due to the way it often uses zip.
Update to latest setuptools. Latest setuptools will always try to run tests.
This can cause some very vague errors. We now need to fix all packages where we do not
invoke the correct test runner.
See http://nixos.org/nixpkgs/manual/#sec-package-naming
I've added an alias for multipath_tools to make sure that we don't break
existing configurations referencing the old name.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
