On Sat, 29 Sep 2018 06:30:39 +0100 Chris Jones
<jon...@hep.phy.cam.ac.uk> wrote:
> > On 28 Sep 2018, at 10:55 pm, Perry E. Metzger
> > <pmetz...@macports.org> wrote:
> >
> > On Sat, 29 Sep 2018 07:40:22 +1000 Joshua Root <j...@macports.org>
> > wrote:
> >>>> Here's the kludge for a Portfile:
> >>>>
> >>>> if {[vercmp ${xcodeversion} 10.0] >= 0} {
> >>>> destroot.pre_args -UseNewBuildSystem=NO
> >>>> }
> >>>>
> >>>> Arguably, one should do "= 0" rather than ">=" since the bug
> >>>> may be fixed in Xcode 10.whatever.
> >>>
> >>> That would depend on whether you are correct or not in saying
> >>> its a bug in Xcode. If it isn’t making it an equality will just
> >>> mean the ports start failing again when Xcode is updated.
> >>
> >> The header search strategy change is not a bug. The failure when
> >> building and destrooting separately is definitely a bug, and can
> >> be reproduced even with a new project using the Cocoa App
> >> template.
> >
> > BTW, that's why I set this up only to do the UseNewBuildSystem=NO
> > during destroot and not during build. Fixing the search stuff for
> > build seems sane.
>
> Hmm... is it really ok to switch the build settings between build
> and destroot ? I would have guessed it better to be consistent for
> all phases.
It seems to work just fine, and it means that when we can remove the
kludge from the destroot phase, which is the only place the bug hits
us, we'll know the rest is also working.
--
Perry E. Metzger pmetz...@macports.org
