Re: A standard patch rule for our rules
Ben Finney <ben+debian@benfinney.id.au> writes:
> Charles Plessy <plessy@debian.org> writes:
>> Unfortunately, it seems that with quilt, it is better ot use
>> $(QUILT_STAMPFN) in order to avoid a target to become phony.
> What's wrong with having a phony target? We already have many of them,
> and a standard way of dealing with them: as dependencies of the
> ‘.PHONY’ target.
> It seems to me that the whole point of adding ‘patch’ as a (phony)
> target is to allow a dependency on that target, instead of something
> patch-system-specific. What is your reasoning for wanting to diverge
> from that?
I think Charles may be referring to the need to use $(QUILT_STAMPFN)
rather than patch as a dependency of a build-stamp target, since
otherwise the dependency on patch forces build-stamp to be always
considered out of date even if the stamp file is up-to-date. This has
caused problems for a few packages where the build target was run
repeatedly in ways that it wasn't intended to. Depending on
$(QUILT_STAMPFN) instead is more reliable.
I'm not entirely sure how that relates to the topic of conversation in
this thread, though. It doesn't affect the presence of the patch target
in debian/rules.
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: