Efraim Flashner <efr...@flashner.co.il> skribis: > On Thu, 12 Nov 2015 14:05:31 +0100 > l...@gnu.org (Ludovic Courtès) wrote: > >> Efraim Flashner <efr...@flashner.co.il> skribis:
[...] >> > ##Failure Location unknown## : Error >> > Test name: N5aria224LpdMessageDispatcherTestE::testSendMessage >> > uncaught exception of type N5aria29DlAbortExE >> > - Failed to set a socket option, cause: No such device >> >> Could you check what this test does? The “No such device” is likely due >> to our build environment not corresponding to the assumptions made by >> the test wrt. available networking capabilities. > > here's the failing line: > std::shared_ptr<SocketCore> recvsock(new SocketCore(SOCK_DGRAM)); How do you know it’s this line? The “Failed to set a socket option” would make me think it’s about a failing setsockopt(2) call, but it’s not supposed to return ENODEV. socket(2) is not supposed to return ENODEV either. Could you keep the failed build directory (with ‘-K’), and from there run the test in ‘strace’ to see exactly what’s failing? (With the caveat that it may succeed outside the build environment; did you try running ‘make check’ from the failed build tree, outside the build environment?) > I'll keep staring at it, and hopefully it'll make sense soon :) It will! :-) Cheers, Ludo’.