[DEP 8] Re: Is it possible to run autopkgtest without a virtual machine ?
Charles Plessy writes ([DEP 8] Re: Is it possible to run autopkgtest without a virtual machine ?): For networking, how about a needs-networking restriction ? I'm definitely in general in favour of expanding the set of restrictions etc. We need to be sure that the semantics are clear though. Would that mean only that the machine needs to have a non-localhost network interface, or also that it can access the global internet somehow (and if so which protocols) ? For protecting the user from side effects of runnign the tests, there is the Restrictions field where one can declare that tests can break the system. Perhaps a milder restriction could be added, for tests that can disturb the system (such as restarting services, etc.), so that other tests can be considered safe for local execution. A large number of regression tests that I know are completely safe, like checking that 2 + 2 returns 4, with of course the limitation that running them exposes to the same possible bugs as for using the program in other contexts. Again, yes. Perhaps you could describe those of your package(s) and the test(s) which would benefit from this ? That would make it easier to think about what exactly the restrictions should be. I think that it would be a great feature to be able to run such tests locally with a simple command that can take advantage of autopkgtest. Yes. This means also shipping them in a binary package, but we have already such packages containing exactly the test data that is needed. No, this won't work because the tests might need stuff from the source package; might need the source package to be built; and we shouldn't clog the archive up with test suites; the build depends would need to be installed anyway. The right answer is simply for such an automatic tool to download the source package. Ian. -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20456.22853.76070.451...@chiark.greenend.org.uk
[DEP 8] Re: Is it possible to run autopkgtest without a virtual machine ?
Le Fri, Jun 22, 2012 at 11:02:04AM +0200, Goswin von Brederlow a écrit : I think test should be run in a container (or VM) but never the local system itself. The absolute minimum would be a chroot but a container would give better control over things like networking. Hello Goswin For networking, how about a needs-networking restriction ? For protecting the user from side effects of runnign the tests, there is the Restrictions field where one can declare that tests can break the system. Perhaps a milder restriction could be added, for tests that can disturb the system (such as restarting services, etc.), so that other tests can be considered safe for local execution. A large number of regression tests that I know are completely safe, like checking that 2 + 2 returns 4, with of course the limitation that running them exposes to the same possible bugs as for using the program in other contexts. I think that it would be a great feature to be able to run such tests locally with a simple command that can take advantage of autopkgtest. This means also shipping them in a binary package, but we have already such packages containing exactly the test data that is needed. Have a nice day -- Charles Plessy Debian Med packaging team, http://www.debian.org/devel/debian-med Tsurumi, Kanagawa, Japan -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20120623032924.gb15...@falafel.plessy.net
