On Mon, May 19, 2008 at 3:20 PM, Thiemo Seufer <[EMAIL PROTECTED]> wrote: > Zack Weinberg wrote: >> http://lists.debian.org/debian-devel/2006/10/msg01278.html >> -- which is inconclusive, but seems to be heading in the direction >> that network-dependent tests are ok as long as the build doesn't fail >> if the network is unavailable. Which is what I would like to do, if I >> had a sane way to determine whether network tests can be run. > > You can run the tests, record their result in the build log, and > continue to build the package regardless of the outcome.
I don't think that's any better than skipping the tests. The build *should* fail if the networking functions of the program are actually broken. > I don't know of a single magic command which would test all potential > network failure modes. I'm currently thinking I should write a little C program that exercises the same parts of the socket API that the server itself uses - but not knowing what the setup is on the mips buildds, I don't know how to verify that that program fails when it needs to... zw -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
