Re: Failure to build www/firefox

[maya%NetBSD.org@localhost]

On Fri, May 08, 2020 at 07:41:06AM +0200, Thomas Klausner wrote:
> On Thu, May 07, 2020 at 09:06:12PM +0200, Martin Husemann wrote:
> > On Thu, May 07, 2020 at 07:55:20PM +0100, Mike Pumford wrote:
> > > Just found this change to clang10 in the right timeframe that looks like it
> > > might impact which libstdc++ we are getting:
> > > 
> > > http://cvsweb.netbsd.org/bsdweb.cgi/pkgsrc/lang/clang/Makefile.common.diff?r1=1.42&r2=1.43&only_with_tag=MAIN
> > > 
> > > Just kicked off a build with that change backed out and see if it helps.
> > 
> > I can tell you without waiting for the build results:
> > 
> > with r1.42 I get SUBST_CLASES libcxx cmake, with r1.42 it is only cmake.
> > 
> > This change is broken and we need to rebuild clang as well.
> 
> Even with the fixed clang, I still see:
> 
> cargo:warning=/scratch/www/firefox/work/build/dist/include/mozilla/Assertions.h:425:23: error: no template named 'is_array_v' in namespace 'std';
> did you mean 'is_array'?
> cargo:warning=  static_assert(!std::is_array_v<ValueT>,
> cargo:warning=                 ~~~~~^~~~~~~~~~
> cargo:warning=                      is_array

This seems to be a c++17 thing.

I wonder if we need to drop our own USE_LANGUAGES=c++14 to c++ and
leave upstream to choose correct values for every file.