Re: [PATCH v6 2/3] Makefile: add Perl runtime prefix support
- From
Daniel Jacques <dnj@google.com>
- Date
- Mar 19, 2018, 17:21 UTC
- Message-ID
- <CAD1RUU9R5xfGghQP-A=yU0RkmVDaS0hPGMHVieCyb-DRg97SnQ@mail.gmail.com>
- In-Reply-To
- <xmqqbmfkvvhc.fsf@gitster-ct.c.googlers.com>
On Mon, Mar 19, 2018 at 1:14 PM Junio C Hamano <gitster@pobox.com> wrote:
> > +# RUNTIME_PREFIX's resolution logic requires resource paths to be
expressed
> > +# relative to each other and share an installation path. > > +# > > +# This is a dependnecy in:
> dependency?
Oops, this is the second typo that has been pointed out. I'll release one last series after a small review period with these fixed.
Show 7 quoted lines
> > +# - Git's binary RUNTIME_PREFIX logic in (see "exec_cmd.c"). > > +# - The runtime prefix Perl header (see > > +# "perl/header_templates/runtime_prefix.template.pl"). > > +ifdef RUNTIME_PREFIX > > + > > +ifneq ($(filter /%,$(firstword $(gitexecdir_relative))),) > > +$(error RUNTIME_PREFIX requires a relative gitexecdir, not:
$(gitexecdir))
> > +endif
Show 5 quoted lines
> I see Dscho is CC'ed so I won't worry about "is there a more > portable test than 'the path begins with a slash' to see if a path > is relative, or is this good enough even for Windows in the context > of this patch?". It won't be a show-stopper issue as long as we do > not error out with false positive, though ;-).
OK sounds good! There are other places in the Makefile that use this method for this purpose, so hopefully the worst-case is that this is no more broken than they are.