Cucumber / RSpec Story Runner

Has anyone tried to get Cucumber (or it’s predecessor: the RSpec story
runner) to work with IronRuby? I have not had any luck thus far (I am
extremely new to Ruby), and I would really like to be able to use
Cucumber to write executable feature documentation for my .NET code.

I copied it under the “lib” directory, as well as several of its
dependencies, but when I try to do “require
‘cucumber-0.1.7/lib/cucumber’”, ir (interactive ruby) fails with a stack
overflow. The same thing happens when I try to require some its
dependencies manually (eg. “require ‘hoe-1.8.0/lib/hoe’” and “require
‘rake-0.8.3/lib/rake’”), but some work just fine (eg. “require
‘polyglot-0.2.3/lib/polyglot’” and “require
‘treetop-1.2.4/lib/treetop’”).

I also tried to get gem working to aide me in this process, but I had
problems with that too. Has anyone got gem working with IronRuby?

Thanks in advance!
Pat Gannon

If you explicitly set the GEM_PATH before requiring gems, you should be
able to use gems that are already present. I haven’t tried any other
gem operations.

I assume you’re running with the latest source?

Hi,

To get rspec to work you will need to modify the actual Ironruby
source. I’ve raised bugs to get the changes required fixed, however I
did this based on 1.1.4, I think Cucumber will have a whole load more
bugs attached as it has more dependencies.

I know for a fact you won’t be able to run Cucumber unmodified due to
existing bugs around gems. (for example, gem needing .rb at the end
of the file to be able to find it)

However, I haven’t had chance to take a closer look.

Ben


Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core


Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core


Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core


Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core


Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core


Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core


Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core


Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core


Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core


Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core


Ironruby-core mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/ironruby-core

I wasn’t really sure where to start.

Running a rake compile leaves the system alone, it’s actually when you
open the projects in visual studio (which I needed to do earlier to
compile a release build).

Note: I have my code in c:\dev\ironruby on this box. there’s no trunk
subfolder, so this may have something to do with all those …
…'s that are in the csproj files

The security warning dialog comes up telling me it may be unsafe, and
asking to load the project for browsing, or to load normally.

I tell it to load the project normally (I trust you guys, awwww), and
it then creates these folders. Note that the only thing that was
present previously was my c:\dev

c:\build\release
c:\Merlin\Main\Bin\FxCop
c:\dev\Bin\FxCop

Note: At this point, all I’ve done is tell VS to load
microsoft.scripting.core.dll
It then prompts me again if I trust it to load
microsoft.scripting.extensionattribute.dll, which doesn’t create any
extra directories that I can find

This is not something I realised visual studio could actually even do,
so perhaps I’m not the right person to provide a patch for it…

Moving on, I now switch to release config, and hit compile. This now
works with no additional hackery in svn r168 (Thanks John!)

However, I now also have

c:\Tools\Nessie\Nessie\bin\IronRuby.dll
c:\build\debug

IIRC compiling from VS in debug mode produces even more littering too.

Thanks, Orion

Sure. Here’s the entire contents of the ruby file

$LOAD_PATH << ‘c:/dev/rspec/lib’

require ‘spec’

module ActiveSupport; end # workaround ironruby defined? bug

class Vehicle
def initialize(people)
@people = people
end
attr_accessor :people
end

describe Vehicle do
it “should assign people using the constructor” do
car = Vehicle.new([‘orion’, ‘john’, ‘jim’])
car.should have(3).people
end
end

Note I’m not using gems or anything else like that. To replicate this
you’d basically do this:

svn checkout ironruby 168 into c:\dev\ironruby

rake compile

git clone rspec straight from github, and stick it in c:\dev\rspec

edit c:\dev\rspec\lib\spec\extensions\main.rb and fix the call to
File.expand_path

spin up a command prompt and set PATH=blahblah so it can see the
ironruby binaries that got build previously

ir rspectest.rb

It appears that rspec has a broken gemspec. It’s not installing
spicycode-rcov, syntax, and possibly more.

JD


From: [email protected]
[[email protected]] On Behalf Of John L. (IRONRUBY)
[[email protected]]
Sent: Tuesday, October 21, 2008 10:40 PM
To: [email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

Orion – can you supply us with your rspectest.rb file?

I was just hacking around with rspec 1.1.9 here, and it’s blowing up
with a bunch of dependencies on startup.

Thanks,
-John

From: [email protected]
[mailto:[email protected]] On Behalf Of Orion E.
Sent: Tuesday, October 21, 2008 7:37 PM
To: [email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

The nested defined? bug is logged here:

http://rubyforge.org/tracker/index.php?func=detail&aid=22503&group_id=4359&atid=16798

The ‘caller’ one is actually 2 bugs, both of which have been logged by
ben hall already

http://rubyforge.org/tracker/?group_id=4359&atid=16798&func=detail&aid=22315
http://rubyforge.org/tracker/?group_id=4359&atid=16798&func=detail&aid=22348

For those interested in running rspec right now, a better fix is to
replace line 26 of main.rb with

args.last[:spec_path] = File.expand_path( caller(0)[1].split(‘:’).first
)

This will only cause rspec to lose the line number - you also have to
run ir with the -D flag, or it doesn’t include the path at all.

As far as my quick-hack investigation shows me, rspec only uses this
data for showing friendly error messages, rather than any core logic, so
perhaps it’s not super-critical.
It would be nice to run it unpatched tho, IronRuby is looking incredibly
close to doing that :slight_smile:

Jim D. wrote:
I’ll add a rake task to compile release. Can you try to isolate those
cases and file bugs? The second one is pretty easy, but I’d like to have
tracking on both of them.

From:
[email protected]mailto:[email protected]
[mailto:[email protected]] On Behalf Of Orion E.
Sent: Tuesday, October 21, 2008 7:04 PM
To: [email protected]mailto:[email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

I just ran rspec 1.1.9 (straight from github) on IronRuby build 167, and
it appears to work (I only have a simple test, but hey).

I only had to make 2 changes:

rspec\lib\spec\extensions\main.rb:26 is

args.last[:spec_path] = File.expand_path(caller(0)[1])
IronRuby still doesn’t seem to handle caller quite the same as MRI. I
have no idea what that code is for, but I replaced it with this as a
quick hack

args.last[:spec_path] = “.”

and it seemed to work.
I encountered another problem using the ‘have’ matcher - rspec does this

if inflector = (defined?(ActiveSupport::Inflector) ?
ActiveSupport::Inflector : (defined?(Inflector) ? Inflector : nil))
IronRuby throws uninitialized constant Object::ActiveSupport, whereas
MRI just returns nil - it looks like IR isn’t handling defined? on
nested things properly yet.

To work around this I just put

module ActiveSupport; end
at the top of my ruby file to get around the nested module issue.

And presto!

C:\development\irtest>ir rspectest.rb

.

Finished in 2.072133 seconds

1 example, 0 failures

C:\development\irtest>ruby rspectest.rb

.

Finished in 0.053 seconds

1 example, 0 failures

The only problem now is that 2.07 seconds is somewhat larger than
0.053… I am using the debug build of ir though, as that’s what
rake:compile seems to give me.

Just playing with VS now to see if I can build a release version and try
that

Ben H. wrote:

Hi,

To get rspec to work you will need to modify the actual Ironruby

source. I’ve raised bugs to get the changes required fixed, however I

did this based on 1.1.4, I think Cucumber will have a whole load more

bugs attached as it has more dependencies.

I know for a fact you won’t be able to run Cucumber unmodified due to

existing bugs around gems. (for example, gem needing .rb at the end

of the file to be able to find it)

However, I haven’t had chance to take a closer look.

Ben

On Fri, Oct 17, 2008 at 2:53 PM, Curt H.
[email protected]mailto:[email protected] wrote:

If you explicitly set the GEM_PATH before requiring gems, you should be
able to use gems that are already present. I haven’t tried any other
gem operations.

I assume you’re running with the latest source?

John, for the record, you have to add github as a source for
spicycode-rcov (gem sources -a http://gems.github.com).

JD


From: [email protected]
[[email protected]] On Behalf Of Jim D.
[[email protected]]
Sent: Tuesday, October 21, 2008 11:10 PM
To: [email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

Here’s another test if you are using gems. It should work with ir
test_spec.rb

JD


From: [email protected]
[[email protected]] On Behalf Of Orion E.
[[email protected]]
Sent: Tuesday, October 21, 2008 11:07 PM
To: [email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

Sure. Here’s the entire contents of the ruby file

$LOAD_PATH << ‘c:/dev/rspec/lib’

require ‘spec’

module ActiveSupport; end # workaround ironruby defined? bug

class Vehicle
def initialize(people)
@people = people
end
attr_accessor :people
end

describe Vehicle do
it “should assign people using the constructor” do
car = Vehicle.new([‘orion’, ‘john’, ‘jim’])
car.should have(3).people
end
end

Note I’m not using gems or anything else like that. To replicate this
you’d basically do this:

svn checkout ironruby 168 into c:\dev\ironruby

rake compile

git clone rspec straight from github, and stick it in c:\dev\rspec

edit c:\dev\rspec\lib\spec\extensions\main.rb and fix the call to
File.expand_path

spin up a command prompt and set PATH=blahblah so it can see the
ironruby binaries that got build previously

ir rspectest.rb

On 22/10/2008, at 6:40 PM, John L. (IRONRUBY) wrote:

Orion – can you supply us with your rspectest.rb file?

I was just hacking around with rspec 1.1.9 here, and it’s blowing up
with a bunch of dependencies on startup.

Thanks,
-John

From:
[email protected]mailto:[email protected]
[mailto:[email protected]] On Behalf Of Orion E.
Sent: Tuesday, October 21, 2008 7:37 PM
To: [email protected]mailto:[email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

The nested defined? bug is logged here:

http://rubyforge.org/tracker/index.php?func=detail&aid=22503&group_id=4359&atid=16798

The ‘caller’ one is actually 2 bugs, both of which have been logged by
ben hall already

http://rubyforge.org/tracker/?group_id=4359&atid=16798&func=detail&aid=22315
http://rubyforge.org/tracker/?group_id=4359&atid=16798&func=detail&aid=22348

For those interested in running rspec right now, a better fix is to
replace line 26 of main.rb with

args.last[:spec_path] = File.expand_path( caller(0)[1].split(‘:’).first
)

This will only cause rspec to lose the line number - you also have to
run ir with the -D flag, or it doesn’t include the path at all.

As far as my quick-hack investigation shows me, rspec only uses this
data for showing friendly error messages, rather than any core logic, so
perhaps it’s not super-critical.
It would be nice to run it unpatched tho, IronRuby is looking incredibly
close to doing that :slight_smile:

Jim D. wrote:
I’ll add a rake task to compile release. Can you try to isolate those
cases and file bugs? The second one is pretty easy, but I’d like to have
tracking on both of them.

From:
[email protected]mailto:[email protected]
[mailto:[email protected]] On Behalf Of Orion E.
Sent: Tuesday, October 21, 2008 7:04 PM
To: [email protected]mailto:[email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

I just ran rspec 1.1.9 (straight from github) on IronRuby build 167, and
it appears to work (I only have a simple test, but hey).

I only had to make 2 changes:

rspec\lib\spec\extensions\main.rb:26 is

args.last[:spec_path] = File.expand_path(caller(0)[1])

IronRuby still doesn’t seem to handle caller quite the same as MRI. I
have no idea what that code is for, but I replaced it with this as a
quick hack

args.last[:spec_path] = “.”

and it seemed to work.

I encountered another problem using the ‘have’ matcher - rspec does this

if inflector = (defined?(ActiveSupport::Inflector) ?
ActiveSupport::Inflector : (defined?(Inflector) ? Inflector : nil))

IronRuby throws uninitialized constant Object::ActiveSupport, whereas
MRI just returns nil - it looks like IR isn’t handling defined? on
nested things properly yet.

To work around this I just put

module ActiveSupport; end

at the top of my ruby file to get around the nested module issue.

And presto!

C:\development\irtest>ir rspectest.rb

.

Finished in 2.072133 seconds

1 example, 0 failures

C:\development\irtest>ruby rspectest.rb

.

Finished in 0.053 seconds

1 example, 0 failures

The only problem now is that 2.07 seconds is somewhat larger than
0.053… I am using the debug build of ir though, as that’s what
rake:compile seems to give me.

Just playing with VS now to see if I can build a release version and try
that

Ben H. wrote:

Hi,

To get rspec to work you will need to modify the actual Ironruby

source. I’ve raised bugs to get the changes required fixed, however I

did this based on 1.1.4, I think Cucumber will have a whole load more

bugs attached as it has more dependencies.

I know for a fact you won’t be able to run Cucumber unmodified due to

existing bugs around gems. (for example, gem needing .rb at the end

of the file to be able to find it)

However, I haven’t had chance to take a closer look.

Ben

On Fri, Oct 17, 2008 at 2:53 PM, Curt H.
[email protected]mailto:[email protected] wrote:

If you explicitly set the GEM_PATH before requiring gems, you should be
able to use gems that are already present. I haven’t tried any other
gem operations.

I assume you’re running with the latest source?

Here’s another test if you are using gems. It should work with ir
test_spec.rb

JD


From: [email protected]
[[email protected]] On Behalf Of Orion E.
[[email protected]]
Sent: Tuesday, October 21, 2008 11:07 PM
To: [email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

Sure. Here’s the entire contents of the ruby file

$LOAD_PATH << ‘c:/dev/rspec/lib’

require ‘spec’

module ActiveSupport; end # workaround ironruby defined? bug

class Vehicle
def initialize(people)
@people = people
end
attr_accessor :people
end

describe Vehicle do
it “should assign people using the constructor” do
car = Vehicle.new([‘orion’, ‘john’, ‘jim’])
car.should have(3).people
end
end

Note I’m not using gems or anything else like that. To replicate this
you’d basically do this:

svn checkout ironruby 168 into c:\dev\ironruby

rake compile

git clone rspec straight from github, and stick it in c:\dev\rspec

edit c:\dev\rspec\lib\spec\extensions\main.rb and fix the call to
File.expand_path

spin up a command prompt and set PATH=blahblah so it can see the
ironruby binaries that got build previously

ir rspectest.rb

On 22/10/2008, at 6:40 PM, John L. (IRONRUBY) wrote:

Orion – can you supply us with your rspectest.rb file?

I was just hacking around with rspec 1.1.9 here, and it’s blowing up
with a bunch of dependencies on startup.

Thanks,
-John

From:
[email protected]mailto:[email protected]
[mailto:[email protected]] On Behalf Of Orion E.
Sent: Tuesday, October 21, 2008 7:37 PM
To: [email protected]mailto:[email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

The nested defined? bug is logged here:

http://rubyforge.org/tracker/index.php?func=detail&aid=22503&group_id=4359&atid=16798

The ‘caller’ one is actually 2 bugs, both of which have been logged by
ben hall already

http://rubyforge.org/tracker/?group_id=4359&atid=16798&func=detail&aid=22315
http://rubyforge.org/tracker/?group_id=4359&atid=16798&func=detail&aid=22348

For those interested in running rspec right now, a better fix is to
replace line 26 of main.rb with

args.last[:spec_path] = File.expand_path( caller(0)[1].split(‘:’).first
)

This will only cause rspec to lose the line number - you also have to
run ir with the -D flag, or it doesn’t include the path at all.

As far as my quick-hack investigation shows me, rspec only uses this
data for showing friendly error messages, rather than any core logic, so
perhaps it’s not super-critical.
It would be nice to run it unpatched tho, IronRuby is looking incredibly
close to doing that :slight_smile:

Jim D. wrote:
I’ll add a rake task to compile release. Can you try to isolate those
cases and file bugs? The second one is pretty easy, but I’d like to have
tracking on both of them.

From:
[email protected]mailto:[email protected]
[mailto:[email protected]] On Behalf Of Orion E.
Sent: Tuesday, October 21, 2008 7:04 PM
To: [email protected]mailto:[email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

I just ran rspec 1.1.9 (straight from github) on IronRuby build 167, and
it appears to work (I only have a simple test, but hey).

I only had to make 2 changes:

rspec\lib\spec\extensions\main.rb:26 is

args.last[:spec_path] = File.expand_path(caller(0)[1])

IronRuby still doesn’t seem to handle caller quite the same as MRI. I
have no idea what that code is for, but I replaced it with this as a
quick hack

args.last[:spec_path] = “.”

and it seemed to work.

I encountered another problem using the ‘have’ matcher - rspec does this

if inflector = (defined?(ActiveSupport::Inflector) ?
ActiveSupport::Inflector : (defined?(Inflector) ? Inflector : nil))

IronRuby throws uninitialized constant Object::ActiveSupport, whereas
MRI just returns nil - it looks like IR isn’t handling defined? on
nested things properly yet.

To work around this I just put

module ActiveSupport; end

at the top of my ruby file to get around the nested module issue.

And presto!

C:\development\irtest>ir rspectest.rb

.

Finished in 2.072133 seconds

1 example, 0 failures

C:\development\irtest>ruby rspectest.rb

.

Finished in 0.053 seconds

1 example, 0 failures

The only problem now is that 2.07 seconds is somewhat larger than
0.053… I am using the debug build of ir though, as that’s what
rake:compile seems to give me.

Just playing with VS now to see if I can build a release version and try
that

Ben H. wrote:

Hi,

To get rspec to work you will need to modify the actual Ironruby

source. I’ve raised bugs to get the changes required fixed, however I

did this based on 1.1.4, I think Cucumber will have a whole load more

bugs attached as it has more dependencies.

I know for a fact you won’t be able to run Cucumber unmodified due to

existing bugs around gems. (for example, gem needing .rb at the end

of the file to be able to find it)

However, I haven’t had chance to take a closer look.

Ben

On Fri, Oct 17, 2008 at 2:53 PM, Curt H.
[email protected]mailto:[email protected] wrote:

If you explicitly set the GEM_PATH before requiring gems, you should be
able to use gems that are already present. I haven’t tried any other
gem operations.

I assume you’re running with the latest source?

Great - I have it working now on my machine.

I’m seeing 0.45s or so as the net time (excluding startup) for the tests
if I run using the -X:Interpret flag. Can you try running with that flag
enabled on your machine and let me know how quickly it runs?

Thanks,
-John

From: [email protected]
[mailto:[email protected]] On Behalf Of Orion E.
Sent: Tuesday, October 21, 2008 11:07 PM
To: [email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

Sure. Here’s the entire contents of the ruby file

$LOAD_PATH << ‘c:/dev/rspec/lib’

require ‘spec’

module ActiveSupport; end # workaround ironruby defined? bug

class Vehicle
def initialize(people)
@people = people
end
attr_accessor :people
end

describe Vehicle do
it “should assign people using the constructor” do
car = Vehicle.new([‘orion’, ‘john’, ‘jim’])
car.should have(3).people
end
end

Note I’m not using gems or anything else like that. To replicate this
you’d basically do this:

svn checkout ironruby 168 into c:\dev\ironruby

rake compile

git clone rspec straight from github, and stick it in c:\dev\rspec

edit c:\dev\rspec\lib\spec\extensions\main.rb and fix the call to
File.expand_path

spin up a command prompt and set PATH=blahblah so it can see the
ironruby binaries that got build previously

ir rspectest.rb

On 22/10/2008, at 6:40 PM, John L. (IRONRUBY) wrote:

Orion - can you supply us with your rspectest.rb file?

I was just hacking around with rspec 1.1.9 here, and it’s blowing up
with a bunch of dependencies on startup.

Thanks,
-John

From:
[email protected]mailto:[email protected]
[mailto:[email protected]] On Behalf Of Orion E.
Sent: Tuesday, October 21, 2008 7:37 PM
To: [email protected]mailto:[email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

The nested defined? bug is logged here:

http://rubyforge.org/tracker/index.php?func=detail&aid=22503&group_id=4359&atid=16798

The ‘caller’ one is actually 2 bugs, both of which have been logged by
ben hall already

http://rubyforge.org/tracker/?group_id=4359&atid=16798&func=detail&aid=22315
http://rubyforge.org/tracker/?group_id=4359&atid=16798&func=detail&aid=22348

For those interested in running rspec right now, a better fix is to
replace line 26 of main.rb with

args.last[:spec_path] = File.expand_path( caller(0)[1].split(‘:’).first
)

This will only cause rspec to lose the line number - you also have to
run ir with the -D flag, or it doesn’t include the path at all.

As far as my quick-hack investigation shows me, rspec only uses this
data for showing friendly error messages, rather than any core logic, so
perhaps it’s not super-critical.
It would be nice to run it unpatched tho, IronRuby is looking incredibly
close to doing that :slight_smile:

Jim D. wrote:
I’ll add a rake task to compile release. Can you try to isolate those
cases and file bugs? The second one is pretty easy, but I’d like to have
tracking on both of them.

From:
[email protected]mailto:[email protected]
[mailto:[email protected]] On Behalf Of Orion E.
Sent: Tuesday, October 21, 2008 7:04 PM
To: [email protected]mailto:[email protected]
Subject: Re: [Ironruby-core] Cucumber / RSpec Story Runner

I just ran rspec 1.1.9 (straight from github) on IronRuby build 167, and
it appears to work (I only have a simple test, but hey).

I only had to make 2 changes:

rspec\lib\spec\extensions\main.rb:26 is

args.last[:spec_path] = File.expand_path(caller(0)[1])
IronRuby still doesn’t seem to handle caller quite the same as MRI. I
have no idea what that code is for, but I replaced it with this as a
quick hack

args.last[:spec_path] = “.”

and it seemed to work.
I encountered another problem using the ‘have’ matcher - rspec does this

if inflector = (defined?(ActiveSupport::Inflector) ?
ActiveSupport::Inflector : (defined?(Inflector) ? Inflector : nil))
IronRuby throws uninitialized constant Object::ActiveSupport, whereas
MRI just returns nil - it looks like IR isn’t handling defined? on
nested things properly yet.

To work around this I just put

module ActiveSupport; end
at the top of my ruby file to get around the nested module issue.

And presto!

C:\development\irtest>ir rspectest.rb

.

Finished in 2.072133 seconds

1 example, 0 failures

C:\development\irtest>ruby rspectest.rb

.

Finished in 0.053 seconds

1 example, 0 failures

The only problem now is that 2.07 seconds is somewhat larger than
0.053… I am using the debug build of ir though, as that’s what
rake:compile seems to give me.

Just playing with VS now to see if I can build a release version and try
that

Ben H. wrote:

Hi,

To get rspec to work you will need to modify the actual Ironruby

source. I’ve raised bugs to get the changes required fixed, however I

did this based on 1.1.4, I think Cucumber will have a whole load more

bugs attached as it has more dependencies.

I know for a fact you won’t be able to run Cucumber unmodified due to

existing bugs around gems. (for example, gem needing .rb at the end

of the file to be able to find it)

However, I haven’t had chance to take a closer look.

Ben

On Fri, Oct 17, 2008 at 2:53 PM, Curt H.
[email protected]mailto:[email protected] wrote:

If you explicitly set the GEM_PATH before requiring gems, you should be
able to use gems that are already present. I haven’t tried any other
gem operations.

I assume you’re running with the latest source?