Fix for bundler path issue in Cucumber Rake task #386

[y-higuchi]

This is a fix for #386.

It changes the method to obtain the bundle command name from
Gem.default_exec_format % 'bundle' to File.basename( Gem.bin_path('bundler', 'bundle') ) when building the command line in Cucumber Rake task.

[os97673]

Could you please extract this code into separate method. To remove code duplication and make this code more readable.

[y-higuchi]

I've extracted the duplicate as ForkedCucumberRunner#bundle_cmd method. d08dc89

[os97673]

It looks like jruby is also affected by the problem, and the fix works for it too.

[y-higuchi]

I've been checking the rubygems source code, and found out that this Gem.bin_path approach will not work on an environment, where bundle command has suffixes. (original issue #324 situation)

---

Gem.bin_path approach problem detail

Gem.bin_path 'bundler', 'bundle' is implemented by simply appending the second argument to the bin directory of the bundler gem package, thus File.basename( Gem.bin_path 'bundler', 'bundle' ) will always result in 'bundle'.
definition of Gem.bin_path
definition of spec.bin_file called inside Gem.bin_path

---

As an alternative approach, I found an interesting comment in Bundler Issue about speeding up bundle exec.
According to the bundler developer's comment, bundle exec foo is equivalent to ruby -rbundler/setup foo.
This approach only requires the path to the bundler gem resides, and eliminates the need to determine the bundler command name.

I have added an commit for this alternative approach.
I've tested on Ubuntu environment, where ruby1.8 is suffixed and bundler is not suffixed.
I haven't been able to test this on a Windows environment yet, since I won't have access to them until sometime next week.

[os97673]

Hm, I've re-read #324 and I'm not sure it fix a real problem :( At least when I've built ruby with --program-prefix and --program-suffix and than installed bundler, I still have bundle binary, not bundle.
So, I'd say we just need to back out the changes for #324.
@mattwynne @y-higuchi what do you think about this?

[y-higuchi]

Reverting #324 will fix the problem I'm facing, so that will be fine for me.
One thing I'm concerned is that, the person who reported #324 seems to be one of the top contributers of rubygems so there may be some circumstances, that we're not aware of, where bundler installed from gems will have suffixes.

[os97673]

I agree there is a possibility to break something (again) but since #324 was integrated into 1.2.2 and it breaks more common situations (such as jruby and mri installed using apt-get on ubuntu) I think it is safer to simply revert the changes and allow the contributor or someone else to re-fix the problem with possible prefix/suffix of bundler's binary.

[os97673]

@y-higuchi thank you very match for your deep evaluation of the problem it helps us a lot.
And sorry that we cause such a headache with 1.2.2, I've just released 1.2.3 hope it will be a pain relief for you :)

[y-higuchi]

@os97673 No problem. It's nothing compared to all the benefit this project has provided me.
I've confirmed that everything now works on my project environment using 1.2.3.
Thank you for the quick release.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.