On Tue, 6 Nov 2007 21:36:31 +0000 Mark Brown <broonie@sirena.org.uk> wrote: > On Tue, Nov 06, 2007 at 08:54:29PM +0000, Neil Williams wrote: > > > The size of the test suite is completely irrelevant. Executing a single > > compiled binary will cause any cross build to break, even if it is a > > 1kb no-op. > > Attempting to execute a binary not supported by the host will obviously > fail but that is not going to cause the package build to fail > automatically. A lot of what you've been saying appears to be > predicated on the assumption that it will. In which circumstances would it not cause a failure? dpkg-buildpackage will fail if any command cannot be executed, whether that is because the file or directory does not exist or whether it was because the file was not executable or the wrong architecture. Whether the executable is called by sh, make, libtool or by some other means, the build will fail. It's not an assumption, it is the (limited) evidence of the builds performed so far. I suppose it would be possible to test for the existence of the executable but you could only check for the right architecture by build depending on 'file' and I don't know of many test suites that go to those lengths. It's easier simply not to run the test suite when cross compiling. -- Neil Williams ============= http://www.data-freedom.org/ http://www.nosoftwarepatents.com/ http://www.linux.codehelp.co.uk/
Attachment:
pgprOkcto444h.pgp
Description: PGP signature