On Thu, Jul 01, 2021 at 06:23:52PM -0400, Jeff King wrote:
> On Tue, Jun 29, 2021 at 01:12:57PM +0200, Ævar Arnfjörð Bjarmason wrote:
>
> > diff --git a/Makefile b/Makefile
> > index faa8900097..2e3b257164 100644
> > --- a/Makefile
> > +++ b/Makefile
> > @@ -2737,10 +2737,12 @@ tags: FORCE
> > $(FIND_SOURCE_FILES) | xargs ctags -a -o tags+ && \
> > mv tags+ tags
> >
> > +cscope.out:
> > + $(QUIET_GEN)$(RM) cscope.out && \
> > + $(FIND_SOURCE_FILES) | xargs cscope -f$@ -b
> > +
> > .PHONY: cscope
> > -cscope:
> > - $(QUIET_GEN)$(RM) cscope* && \
> > - $(FIND_SOURCE_FILES) | xargs cscope -b
> > +cscope: cscope.out
>
> Doesn't this break subsequent runs after the first generation? With a
> phony "cscope" target, "make cscope" will always run the command, even
> if it's not necessary. But with a real "cscope.out" target but not
> dependencies, it will _never_ run it, even if one of the files changed.
>
> E.g., with your patch:
>
> $ make cscope.out
> GEN cscope.out
>
> $ make cscope.out
> make: 'cscope.out' is up to date.
>
> $ echo 'void foo(void) { }' >>git.c
> $ make cscope.out
> GIT_VERSION = 2.32.0.96.g5daee1b7bb.dirty
> make: 'cscope.out' is up to date.
Ah, I see it is un-broken in the next commit, which adds actual
dependencies. I think it is OK to have a temporarily-broken state in the
history for something so trivial, but it might be worth mentioning it in
the commit message.
-Peff
