On 28/08/07 at 13:49 +0200, Thijs Kinkhorst wrote: > On Tuesday 28 August 2007 09:36, Bas Zoetekouw wrote: > > You wrote: > > > But most people simply don't know that their rules file corrupts > > > tapes. First thing would be to detect these packages. This could > > > be done in the autobuild procedure done on the debian hosts, e.g. > > > by setting $TAPE to point to some kind of "watchdog file". > > > > It should also be possible to detect with with lintian, I would think. > > Simply checking for a tar invocation without -f option should do the > > trick. > > I'm not sure if such a check would be easy to implement reliably in lintian. > > This really sounds like a check that could be implemented using Mole. That is > (or will be) very well suited to test the entire archive e.g. with such > a "strange" build environment, in the way Harald suggests. From what I have understood so far about Mole (and I might be wrong), Mole is not a silver that will make it dead easy to implement QA tests. It will help QA data storage and data mining by providing a single, central place for all QA test results, and help distributing the load of doing the actual tests. But to do this actual test, using a slightly modified sbuild [the packaged one] sounds like a much better plan. One would have to set the env variable before starting the build, and add a check in sbuild so that if complains if the fake tape has been written to. I can run the test rebuild over all packages if someone provides a working patch against sbuild. Also, can someone give an example package that should fail that test? -- | Lucas Nussbaum | lucas@lucas-nussbaum.net http://www.lucas-nussbaum.net/ | | jabber: lucas@nussbaum.fr GPG: 1024D/023B3F4F |
Attachment:
signature.asc
Description: Digital signature