platform can be of type string so it must be cast into a Gem::Platform

[jaym]

#3553 didn't actually fix anything and made things worse (universal gems no longer work on 32 bit ruby on windows)

It seems that the platform can be of type string, so wrapping in in Gem::Platform is necessary.

sorry for the regression

[segiddins]

Since Gem::Platform.new(platform) is done twice, maybe it should be wrapped in a variable?

[segiddins]

@jdmundrawala maybe it would just be better to do this in the initializer for LazySpecification?

[jaym]

@segiddins I'm a little worried about changing LazySpecification and EnpointSecification to do that.
A similar thing was done in rubygems, and produced a bug like: rubygems/rubygems#1184

[segiddins]

Ah actually you’re right, I made that very mistake this weeked

On Apr 13, 2015, at 10:29 AM, Jay Mundrawala [email protected] wrote:

@segiddins https://github.com/segiddins I'm a little worried about changing LazySpecification and EnpointSecification to do that.
A similar thing was done in rubygems, and produced a bug like: rubygems/rubygems#1184 rubygems/rubygems#1184
—
Reply to this email directly or view it on GitHub #3565 (comment).

[jaym]

So reading this again, maybe the === comparison is all that is needed and the == can be dropped

[indirect]

The vast, vast majority of these comparisons are going to exit on the first three conditions, though, so I think it's a performance loss to create a Platform before we do the compare... can we switch to just the one === and wait to create the Platform until then?

[jaym]

Not really certain why the tests keep failing....

  it "installs the latest version of gxapi_rails", :ruby => "1.9"  do
    install_gemfile <<-G
    source "https://rubygems.org"

    gem "sass-rails"
    gem "rails", "~> 3"
    gem "gxapi_rails"
    G
    expect(out).to include("gxapi_rails 0.0.6")
  end

seems to never return. Tried running locally(on a mac) with rubygems 2.2.3 and ruby 1.9.3 with the following Gemfile

source "https://rubygems.org"

gem "sass-rails"
gem "rails", "~> 3"
gem "gxapi_rails"

never returns. Tried with bundler 1.9.2 and bundler 1.9.3

never in this case means

Fetching gem metadata from https://rubygems.org/...........
Fetching version metadata from https://rubygems.org/..
Resolving dependencies..................................................................................................................................................................................................................................................................................................................................................^C

[segiddins]

@jdmundrawala that would imply that the resolver is either looping infinitely, or it's missing the match that it needs to terminate in a sane amount of time.

[jaym]

bundle install 171.70s user 1.46s system 97% cpu 2:58.17 total
It just took a while

[indirect]

@segiddins that seems like a pretty pathological resolver case, any ideas?

[indirect]

(For the record, 1-8-stable takes 19 seconds to finish resolving that Gemfile, while 1-9-stable takes... much longer.)

[indirect]

@jdmundrawala Thanks for the patch!

[indirect]

Cherry-picked to 1-9-stable as 99a7129 and de368e1.