On Wed, Nov 13, 2024 at 7:01 PM George Koehler <kernigh@gmail.com> wrote:

> > On Sat, Oct 5, 2024 at 10:03 AM Jeremy Evans <jeremy@openbsd.org> wrote:
> >
> > > This allows building qtwebkit the current default ports version of
> > > Ruby.  This is a necessary step to remove Ruby 3.1.
> > >
> > > No RDEP on ruby means this is only a build system change, not a
> > > package change, so no bump is needed.
>
> Cc rsadowski, who is maintainer of x11/qt5/*.
>
> jeremy's parser.rb patch works for me.  I looked at webkit in
> https://repology.org/ and noticed that some Linux distros use a patch
> from upstream webkit (https://commits.webkit.org/219400@main).  I
> swapped in that patch, which also works for me.  With either patch, I
> can package qtwebkit-5.212.0alpha4p13v0 on my amd64 with Ruby 3.3 and
> without Ruby 3.1.
>
> The problem is a change to operator =~ in Ruby 3.2.  If @tokens[@idx]
> was in the wrong class, then "@tokens[@idx] =~ final" was false in 3.1
> but raised an error in 3.2.  jeremy's patch fixes it by defining
> Object#=~ to be false.  Upstream's patch is less complete; it moves up
> 1 of 3 checks for "@tokens[@idx].is_a? Annotation", fixing only 1 of 3
> "=~"s.  The other 2 "=~"s might still raise an error, but they didn't
> break my build.
>
> jeremy's patch also redefines String#=~ in a way that would break the
> match variables like $& and $1, but didn't break my qtwebkit build.
> If we pick jeremy's patch and it builds without the String#=~ part,
> ok gkoehler@
>
> If we pick upstream's patch, ok gkoehler@
>

Thank you very much for testing.  I'll commit upstream's patch.  No point
in deviating unnecessarily from upstream.  I'll see if I can do it via a
distpatch instead of a local patch.

Thanks,
Jeremy