The Umlaut problem: UTF-8 chars get converted to unicode entities

[technicalpickles]

The Umlaut problem: UTF-8 chars get converted to unicode entities

[chastell]

BTW: \303\274 is the proper binary representation of UTF-8 ü:

>> "Markus Hüberman" == "Markus H\303\274berman"
=> true

[technicalpickles]

I'm not super familiar with UTF-8 in general, so I'm not sure I can fix this myself.

I've found some notes about doing a File.open to write as utf-8, but it seems to be Ruby 1.9 specific.

[chastell]

Hm, I’m using Ruby 1.9 and UTF-8 in my summary field and both the gemspec file and the resulting gem’s metadata contain the proper UTF-8 strings – i.e., this bug does not manifest itself in my case.

Markus: I tried your name in my gem and it worked. Did you specify the right Ruby file encoding (my Rakefile’s first line is ‘# encoding: UTF-8’)?

[SFEley]

It would make sense for the generated Rakefile to add # encoding: UTF-8 as its first line. This is harmless in Ruby 1.8 but very important in Ruby 1.9 for consistency.

[danopia]

It probably writes like that under 1.8 because it looks like it uses .inspect to dump the string back out. 1.8 isn't aware about UTF-8, it just sees special chars and escapes them. 1.9 is more intelligent about UTF-8 so it'll output as expected.

To fix this if you apply the following string:
# encoding: utf-8
with the hash at the top of each generated gemspec file it should resolve the issue.
I could be wrong but this works for me locally

sorry just realised SFEley also wrote this. +1 on this working :-)

[amatsuda]

Just sent a pull request that adds the magic comment on top of Rakefile. #165

The original decoded UTF-8 problem still remains though. I guess this is a Rubygems' problem.

[technicalpickles]

I've merged in amatsuda's change. It adds # encoding: utf-8 to new project's Rakefile.

Between that, and generated gemspects including # encoding: utf-8, things should be good. Unfortunately, it's not working on 1.8 and existing projects would need to add that line to their own Rakefile.

[gmallard]

In 1.8 you might be able to cure the 1.8 difficulties with:

$KCODE = "U"

Just a thought.