Improve AppVeyor Build

[mvz]

Summary

Work towards a green build on AppVeyor

Details

• Ensure Ruby version variables match actual versions
• Get ffi to install on older Rubies The FFI gem for mingw32 does not support 1.9.3.
• Fix real test failures

Motivation and Context

AppVeyor builds always fail and hence are worse than useless right now.

Types of changes

• Bug fix (non-breaking change which fixes an issue)

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in a week if no further activity occurs.

[olleolleolle]

Can we rebase/merge master into this?

[mvz]

Yes, I'll do a rebase.

[mvz]

.. maybe even split it up so at least the build runs with the correct ruby versions.

[mvz]

Specs are green 🎉. Now for the features 🚧 👷‍♂️ 👷‍♀️ 🏗️

[olleolleolle]

Congratulations on reaching a milestone on this! Getting specs to green is a summit to enjoy.

Next mountain will be possible to scale now that you’ve seen how high up you are.

(The only way is up.)

[olleolleolle]

@mvz Was thinking of this branch, and how it makes the AppVeyor target list shorter - can THAT part be cherry-picked earlier than we fix all features for Windows support, too?

[mvz]

@olleolleolle I suppose all of it can be merged (once I fix the Travis build), since passing specs is better than failing specs.

[xtrasimplicity]

@mvz Mind if I rebase and take a crack at this?

[mvz]

@xtrasimplicity please go ahead. As far as I remember, the current status is that specs are succeeding and cukes are failing. The main reason for the failing cukes is that Windows won't accept executables that have names without extensions, so all the scenarios that create some bin/foo script fail. The quick fix would be to just add a tag like @needs-unix to those scenarios.

[xtrasimplicity]

Hey All,

I have some spare time this weekend to work on getting Aruba's test suite to work on Windows, and I was thinking I'd create Windows-specific versions of the current Linux-specific tests.

Given the current tests are used as documentation and I don't want to pollute the documentation with almost entirely-duplicated tests, I was thinking I might create a seperate folder under features/ for miscellaneous Windows-specific tests (that don't need to be included as documentation)?

Take the following scenario, for example. I don't think we need to have the exact same scenario duplicated beneath it, using echo %cd% or cd instead of pwd. Should I create a Windows-specific copy of this and save it under, say, features/09_windows_specific/cleanup_working_directory.feature? Or perhaps have a separate file under features/01_getting_started_with_aruba, called cleanup_working_directory_windows.feature?

# features\01_getting_started_with_aruba\cleanup_working_directory.feature
 Scenario: Clean up artifacts and pwd from a previous scenario
    Given a file named "features/cleanup.feature" with:
    """
    Feature: Check
      Scenario: Check #1
        Given a file named "file.txt" with "content"
        And a directory named "dir.d"
        Then a file named "file.txt" should exist
        And a directory named "dir.d" should exist
        When I cd to "dir.d"
        And I run `pwd`
        Then the output should match %r</tmp/aruba/dir.d$>
      Scenario: Check #2
        Then a file named "file.txt" should not exist
        And a directory named "dir.d" should not exist
        When I run `pwd`
        Then the output should match %r</tmp/aruba$>
    """
    When I run `cucumber`
    Then the features should all pass

For scenarios like the following, I would create a @requires-msdos or requires-cmd tag and create a MS-DOS/CMD prompt-specific feature, as it should be included in the documentation.

 @requires-bash
  Scenario: Bash Program run via bash
    Given a file named "features/hello_aruba.feature" with:
    """
    Feature: Getting Started With Aruba
      Scenario: First Run of Command
        Given a file named "cli.sh" with:
        \"\"\"
        echo "Hello, Aruba!"
        \"\"\"
        When I successfully run `bash ./cli.sh`
        Then the output should contain:
        \"\"\"
        Hello, Aruba!
        \"\"\"
    """
    When I successfully run `cucumber`
    Then the features should all pass

So, ultimately, how should I handle features that we don't want to appear in Documentation? I'd rather not omit them completely, as our test coverage on Windows will continue to suffer.

Any suggestions would be greatly appreciated. :)

Thanks!

@aslakhellesoy @olleolleolle @mvz -- Sorry for tagging all of you; just keen to get working on this as soon as I can. 👍

[mvz]

I see you found a solution :-).

[xtrasimplicity]

Yep, I've added it to the 'unpublished' directory and updated the .cucumberproignore file. Happy to restructure this if need be - I'm just eager to get the AppVeyor build to pass. :)

[mvz]

Closing in favor of #894.