On 4/4/18, Georg Faerber <ge...@riseup.net> wrote: >> I've checked ruby-gettext-setup and can say that it doesn't actually >> use anything from gettext gem, only from fast-gettext. So its test >> failures caused by something else, not the new gettext version. > I would be interested in how you came to this conclusion:
I did `fgrep -R require | fgrep gettext` and saw nothing (actually this gem uses rxgettext from ruby-gettext). > ruby-gettext-setup FTBFS now, with errors like > > /usr/lib/ruby/vendor_ruby/gettext/tools/parser/ruby.rb:50:in `gsub': invalid > byte sequence in US-ASCII (ArgumentError) > > See #894829 for details. > /usr/lib/ruby/vendor_ruby/gettext/tools/parser/ruby.rb is UTF-8 file, it contains japanese letters (こんにちは世界), of course these letters can't be processed in C locale. So you either patch it to remove UTF-8 characters or run tests in C.UTF-8. I don't think that the last ruby-gettext update introduced a regression as in the previous version it had a terrible insecure parser for ruby strings: begin; s = eval(s); rescue Exception; end Now it has a proper one.