Re: [PATCH] Update Ruby to 2.3.0 (was Re: Freezing core-updates soon)

[Ben Woodcroft]

On 06/01/16 08:24, Ludovic Courtès wrote:
> Ben Woodcroft <address@hidden> skribis:
>
> > I had some trouble downloading dependencies because of inability to
> > download sources e.g. http://graphviz.org/ seems to be down. Ruby
> > seems to build though, can we just review as normal and then push to
> > core-updates?
> >
> > I did rebuild all the packages in ruby.scm, and out of these about 1/3
> > of packages failed to build, because at least these three packages'
> > tests fail
> > * yard
> > * power_assert
> > * minitest_tu_shim
> >
> > The first two presumably will be fixed upstream soon, but the third
> > hasn't seen a release since Aug 2013, and the homepage listed on
> > rubygems is dead. I guess we just have to encourage the only package
> > we have packaged that uses tu_shim (term-ansicolor) to stop relying on
> > it.
> OK, thanks for testing this thoroughly.  It would be good to fix these 3
> packages sometime before core-updates is merged.
In the attached patches I've disabled the failing test in both yard and 
power_assert. There's no fix for yard in upstream. For power_assert 
there is a fix, but unfortunately it seems Ruby packages cannot be 
patched in the usual way because .gem files are archives of archives so 
the unpacking/patch/repacking fails. I guess we fix that in future.

Turns out fixing minitest_tu_shim (and others) was easier than I thought 
as the problem was some hardcoded ruby 2.2 paths in package definitions. 
I created a new procedure to remove the hard-coding so the update to 2.4 
will work too, WDYT?

After all that, everything in gnu/packages/ruby.scm builds.
> > +       (modify-phases %standard-phases
> > +         (add-before 'configure 'replace-bin-sh
> > +           (lambda _
> > +             (substitute* '("Makefile.in"
> > +                            "ext/pty/pty.c"
> > +                            "io.c"
> > +                            "lib/mkmf.rb"
> > +                            "process.c"
> > +                            "test/rubygems/test_gem_ext_configure_builder.rb"
> > +                            "test/rdoc/test_rdoc_parser.rb"
> > +                            "test/ruby/test_rubyoptions.rb"
> > +                            "test/ruby/test_process.rb"
> > +                            "test/ruby/test_system.rb"
> > +                            "tool/rbinstall.rb")
> > +               (("/bin/sh") (which "sh")))
> > +             (substitute* "ext/fiddle/libffi-3.2.1/configure"
> > +               (("SHELL = /bin/sh")
> > +                (string-append "SHELL = " (which "sh"))))
> > +             #t)))))
> > +    (inputs
> > +     `(("readline" ,readline)
> > +       ("openssl" ,openssl)
> > +       ("libffi" ,libffi)
> > +       ("gdbm" ,gdbm)
> > +       ("zlib" ,zlib)))
> > +    (native-search-paths
> > +     (list (search-path-specification
> > +            (variable "GEM_PATH")
> > +            (files (list (string-append "lib/ruby/gems/"
> > +                                        (version-major+minor version)
> > +                                        ".0"))))))
> This seems to be shared with 2.2.  Is it possible to avoid duplication
> using something like:
>
>    (define ruby
>      (package
>        (inherit ruby-2.2)
>        (version "2.3.4")
>        ;; …
>        ))
>
> ?
[..]

Yes, I did this but forgot to delete some/most duplication. Now after 
the final patch ruby-2.2 inherits from ruby (not vice-versa as you 
suggest) as this was the previous convention in ruby.scm. I checked that 
the hashes in the store names for ruby 2.2, 2.1 and 1.8 did not change 
after applying the patch.

Better? Thanks for the review.
ben

0001-gnu-ruby-power-assert-Disable-failing-test.patch
Description: Text Data

0002-gnu-ruby-yard-Disable-failing-test.patch
Description: Text Data

0003-ruby-Abstract-out-path-to-GEM_HOME.patch
Description: Text Data

0004-gnu-ruby-Update-to-2.3.0.patch
Description: Text Data