Issues with release script

[justin808]

I documented the problem here.

Essentially, the release script works, but there's an error.

[justin808]

Recommendation is to switch to using the bundler rake tasks:

• https://www.schneems.com/blogs/2016-03-18-bundler-release-tasks/
• gem release returns 1 (error) always after a gem yank rubygems/rubygems.org#1577 (comment)

Any volunteers to help me with this one?

[isolo]

I'll do it

[isolo]

@justin808 I've debugged gem-release a little. I think I found the reason why it tries to release a version twice. gem-release goes through every gemspec file and applies a command to each of them. For example, for react_on_rails that method returns something like that:
[".", "spec/dummy/client/node_modules/react-on-rails/gen-examples/examples/basic/client/node_modules/react-on-rails", "spec/dummy/client/node_modules/react-on-rails/gen-examples/examples/basic-server-rendering/client/node_modules/react-on-rails", ...]
As you can see, that gem tries to apply a command to all of this gemspecs. Consequently, when it release a version for current directory, it goes to the next and fails with Repushing of gem versions is not allowed.
I'll create a pull request tomorrow.

[justin808]

@isolo can you maybe submit a PR to the gem-release, or maybe at least copy this comment there: svenfuchs/gem-release#63

[dwradcliffe]

Great debugging @isolo!

[isolo]

Fixed that with #778

[justin808]

Great work @isolo! Moving the discuss to the PR!

[isolo]

Yeah, If we delete ./spec/dummy/client/node_modules/react-on-rails/react_on_rails.gemspec I think it solves the problem (the other occurrences are deleted by Rake::Task["examples:clobber"].invoke), but it sounds a little bit shaky. gem-release does not look like a maintainable project, so we can assume (or freeze the version) that its behaviour wouldn't change and anyway we have a dependency on it (bumping version).
I think any of these variants are acceptable. What do you think @justin808 ? It's not hard to make any of these variants workable.

[justin808]

Longer term, we should follow the schneems article. I'd feel more comfortable keeping the script as-is with the addition of deleting the extra gemspecs, maybe via a find command that excludes the top level one.