I, too, think this would be a valuable addition.
On Tue, Feb 09, 2010 at 02:38:34PM -0500, Joey Hess wrote: > If I did this, I would need to also make override_dh_strip to be > skipped when DEB_BUILD_OPTIONS=nostrip. Yes, would also be nice. > One reason to dislike this is it would mean redundant tests in > dh that'd have to be maintained in parallel with the tests in the > commands. Agree that it would take work to implement, though I would think that the tests could be centralized and used by both dh and dh_auto_test. Alternatively: if the logic was all in dh (to skip both dh_auto_test and override_dh_auto_test), then it would not need to be in dh_auto_test at all. > Perhaps a better reason to dislike this is that it violates least > suprise when rules file refactoring. One can move anything into an > override target, even if it does not really match the overridden > debhelper command, and expect it to be run at the appropriate point. > > Here's an example that would cause the package to FTBFS if nocheck > were tested. > > override_dh_auto_test: > # the test suite does not 100% pass at present, > # but the output is useful documentation for users > (dh_auto_test; echo $?) > test-results > > override_dh_install: > dh_install test-results /usr/share/doc/alien/ One could also argue that THIS EXAMPLE violates the principle of least surprise -- in two ways. First, from the builder's point of view: I might set "nocheck" (say, for resource constraint considerations) but have the checks run anyway. Secondly, from a short-on-time debian maintainer's [1] point of view: I might initially use the default dh_auto_test, but then find I need to do something a bit special and write the override -- and be surprised that I've unwittingly broken the "nocheck" feature. Perhaps unsurprisingly, I believe my examples are more common than Joey's :-), so I second the original feature request. -Steve [1] This just happened to me, which is why I'm writing to support this feature request.
signature.asc
Description: PGP signature
