Created attachment 380654 [details, diff] Patch for dev-lisp/sbcl ebuilds disabling contrib tests unless specifically requested The dev-lisp/sbcl ebuilds always run the tests of the contrib modules, even when FEATURES=test is disabled. One of these tests (belonging to sb-bsd-sockets) fails on systems using SELinux. Now, we could specifically disable that test; but I feel the best solution would be to disable all tests if FEATURES=test is disabled, thereby adhering to established ebuild semantics. The attached patch does this by removing the recipe from the test target in contrib/asdf-module.mk using sed. This solution will avoid bugs like 499220 and 486552 in the future.
Ping. (The patch doesn't apply cleanly anymore, but that's just because the context has changed a bit.) @Mark: I've CC'ed you because of your recent activity on the ebuild; the maintainers seem to be unresponsive.
The upstream thinks that tests Including contrib) should be run every time sbcl is being built. Is it really so important to require tests not to be run with USE=-test? Even at the price of a larger distance from the upstream, and hence extra maintainance work? The patch may stop to apply in the future, etc. I'd prefer to stay as close to the upstream as possible.
(In reply to Andrey Grozin from comment #2) > The upstream thinks that tests Including contrib) should be run every time > sbcl is being built. Is it really so important to require tests not to be > run with USE=-test? Even at the price of a larger distance from the > upstream, and hence extra maintainance work? The patch may stop to apply in > the future, etc. I'd prefer to stay as close to the upstream as possible. I don't care whether those tests are executed unconditionally or not. I just proposed this change because it's usually suggested to run tests conditionally (it might even be tree policy, but I'm not sure about that). However, you might want to consider that there's a reason for this suggestion -- it's quite common that one want to install a package even though some tests fail, see for example the two bugs I mentioned in my inital bug description. If you'd rather continue running the tests unconditionally, it'd be great if you could at least disable the sb-bsd-sockets test that fails on SELinux systems and made me report this bug. I'll check again which specific test that is.
Created attachment 398226 [details, diff] Patch to disable sbcl tests failing on SELinux systems I've checked which sb-bsd-sockets test fail in an SELinux environment. Actually, half of them. Here's a patch to disable them.
Created attachment 398230 [details, diff] Merged patch Alternatively, here's a patch including both the already existing bsd-sockets-test-1.2.7.patch and my new -selinux patch.
*sbcl-1.2.9 (07 Mar 2015) 07 Mar 2015; Andrey Grozin <grozin@gentoo.org> +sbcl-1.2.9.ebuild, +files/bsd-sockets-test-1.2.9.patch: Version bump, commenting out one more test in bsd-sockets (#517004), adding pax-mark in src_install (#517008)