On Thu, 2012-03-08 at 22:22 +0700, Michel Alexandre Salim wrote: > On 03/04/2012 01:13 AM, Sérgio Basto wrote: > > On Sat, 2012-03-03 at 10:24 +0700, Michel Alexandre Salim wrote: > >> The tests involve importing some libraries that fail if DISPLAY > >> is not set, thus the reason for using xvfb-run. The same tests > >> run fine on my local mock set up. > > > > > > But if it fail and we don't know why , we need logs of xvfb-run to > > know whats happen , whats is odd ( for me of course) is this test > > when fail, don't have logs about what happened. > > > > > >> And, to my original hypothesis, the failures are not due to the > >> unterminated processes -- I applied a patch from the trunk > >> sourcecode that fixes that issue, and RPM Fusion's mock still > >> failed. > >> > >> Disabling the tests for now (unless building with --with check, > >> so I can still run the tests manually before submitting an > >> official build) but this is really odd. > > > > Let me know if builds after disable the tests. > > > It does build when the tests are disabled, since %check is actually > performed after installation has succeeded, so the only way for the > build to fail at that point is if the file listings are incomplete.
Maybe is a security issue , a build root is not exactly same as your system, for example in build root uname -a (the kernel) is the host not the guset (build root one) , so some things like run a xserver, could be block by security issues, or xorg-x11 may be not compatible with host kernel etc. So a way to know what errors xvfb-run have, is the way to solve this issue . Best regards, -- Sérgio M. B.
