Created attachment 526142 [details, diff] aubio-0.4.6-test-disablement-hack.patch It seems that =media-libs/aubio-0.4.6 has two problems. First, although the ebuild attempts to explicitly request by command-line argument that tests not run, waf goes ahead and runs them anyhow. Second, the tests want to do GPU things, causing a sandbox violation. In order to prevent this, I implemented a local workaround which I've enclosed. It's pretty kludgey.
Comment on attachment 526142 [details, diff] aubio-0.4.6-test-disablement-hack.patch Not sure what this patch actually does. Could you attach the entire build log to this bug report so that others can see the problem you're seeing?
Created attachment 526148 [details] build.log from my box Here's a build.log. FWIW, of the two problems cited, I am coming to suspect that the sandbox violation is some kind of bug idiosyncratic to my workstation. Or perhaps one of those things that only happens when you emerge -e @world, which I happen to be doing (in which case it will be "eventually ubiquitous" :P). Anyhow I don't see bugzilla getting filled with similar reports and yet I've now found several seemingly harmless packages exhibiting this locally (often when some opencl-enabled executable is loaded). I've been unable to figure out exactly what's going on even though I've chased it at length in gdb. Anyhow: the first problem, that of the tests running when they shouldn't, is almost certainly legit. See waf steps starting around #155. My patch simply strips the tests out of the tree of script dependencies by telling the script to look for tests that aren't there (instead of tests that are). But as I said, that's really kludgey. Someone who understands waf should probably do something more elegant -- it's just a work-around.
Created attachment 526150 [details] reporter_provided-build.log updated build.log with ansi color disabled.
I've changed the summary to reflect the correct ebuild phase. To be clear, the tests are running during src_compile. The ebuild /attempts/ to support testing in src_test and furthermore, to suppress testing in src_compile. But the mechanism it's using apparently doesn't work anymore.
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=4f613c9fe28398d60b4329e2e9e64dd5316579fd commit 4f613c9fe28398d60b4329e2e9e64dd5316579fd Author: Andreas Sturmlechner <asturm@gentoo.org> AuthorDate: 2018-10-01 20:07:08 +0000 Commit: Andreas Sturmlechner <asturm@gentoo.org> CommitDate: 2018-10-01 20:08:19 +0000 media-libs/aubio: Disable tests for real Tests were always built and run regardless. At least disable them if not requested. Bug: https://bugs.gentoo.org/651956 Signed-off-by: Andreas Sturmlechner <asturm@gentoo.org> Package-Manager: Portage-2.3.50, Repoman-2.3.11 media-libs/aubio/aubio-0.4.7.ebuild | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-)
Assuming this is fixed.
