Commits on Jun 30, 2016
  1. @hsbt
Commits on May 31, 2016
  1. @schneems

    Merge pull request #488 from heroku/schneems/fuuuuuuuuuuuu-ruby-versi…

    …on-specifiers
    
    Revert "Revert "Bump bundler 1.12.5""
    schneems committed May 31, 2016
Commits on May 27, 2016
  1. @schneems

    Revert "Revert "Bump bundler 1.12.5""

    This reverts commit 088683a.
    
    This PR bumps up the version of bundler to 1.12.5 which allows us to use Ruby version specifiers.
    
    Due to change in behavior in bundler bundler/bundler#4621 we must now detect the when patchlevel is not needed and strip it everywhere we were previously using a version.
    
    Right now there is no way to know when a bundle platform failed with an invalid output, it doesn't do Ruby version validations so if you put "2.3.x" in your Gemfile, then it will return `ruby 2.3.x` back out at you which isn't a valid version and we can't download it. 
    
    I'm going to update the "specify a ruby version" docs to include the ruby version specifiers. I've made this error when a build fails a bit cleaner. We're now only warning about Ruby 2.1 when they're trying to use Ruby 2.1. Also i'm identifying the curl output as "debug information" so hopefully fewer people will think that there is some really strange buildpack error and instead focus on our well written message with links. 
    
    I would like to call out that for failed downloads people should try to re-deploy before declaring it a lost cause, but that's a bit too verbose for writing in an encyclopedic way inside the error output. Think most people will retry anyway. I'm also bumping up the retries to 5, because why not? It might mean that people deploying bad versions of Ruby get delayed by an extra 2 seconds, but hopefully it will cut down on the number of tickets that are solved by people re-running the deploy.
    schneems committed May 26, 2016
Commits on May 26, 2016
  1. @schneems

    Revert "Bump bundler 1.12.5"

    This reverts commit 377d853.
    
    Caused an issue internal ticket reference 364182
    
    Gemfile:
    
    ```
    ruby "2.3.0"
    ```
    
    Gemfile.lock
    
    ```
    RUBY VERSION
       ruby 2.3.0p0
    
    BUNDLED WITH
       1.12.4
    ```
    
    Results in an invalid download URL
    
    ```
     !     An error occurred while installing Ruby ruby-2.3.0-p0
     !     For supported Ruby versions see https://devcenter.heroku.com/articles/ruby-support#supported-runtimes
     !     Note: Only the most recent version of Ruby 2.1 is supported on Cedar-14
     !     Command: 'set -o pipefail; curl -L --fail --retry 3 --retry-delay 1 --connect-timeout 3 --max-time 30 https://s3-external-1.amazonaws.com/heroku-buildpack-ruby/cedar-14/ruby-2.3.0-p0.tgz -s -o - | tar zxf - ' failed unexpectedly:
     !
     !     gzip: stdin: unexpected end of file
     !     tar: Child returned status 1
     !     tar: Error is not recoverable: exiting now
    ```
    
    The URL should be:
    
    ```
    https://s3-external-1.amazonaws.com/heroku-buildpack-ruby/cedar-14/ruby-2.3.0.tgz
    ```
    
    Without the p0.
    
    Investigating
    schneems committed May 26, 2016
  2. @schneems

    Merge pull request #484 from heroku/schneems/bundler-fuzzy-version

    Bump bundler 1.12.5
    schneems committed May 26, 2016
  3. @schneems

    Bump bundler 1.12.5

    Allows for Ruby version operators such as
    
    ```
    ruby "~> 2.3.1"
    ```
    
    When you bundle with this set a Ruby version is written to the `Gemfile.lock`.
    schneems committed May 14, 2016
  4. @schneems

    Update parallel_test && rspec retry

    Apparently some rspec behavior changed with regard to the `pending` tag. Commenting out for now
    
    Fixing runner test, The replrunner is returning "loworld" instead of "helloworld"
    schneems committed May 24, 2016
Commits on May 24, 2016
  1. @hsbt
Commits on May 23, 2016
  1. @schneems

    Merge pull request #486 from TheNotary/patch-2

    makes node come in over https rather than http
    schneems committed May 23, 2016
Commits on May 22, 2016
  1. @TheNotary

    makes node come in over https rather than http

    I haven't tested this on Heroku, so be sure to double check nothing odd happens.
    TheNotary committed May 22, 2016
Commits on Apr 25, 2016
  1. @hsbt
Commits on Apr 18, 2016
  1. @schneems

    Merge pull request #478 from znz/fix-typo

    Fix a typo
    schneems committed Apr 18, 2016
Commits on Apr 6, 2016
  1. @hsbt
Commits on Mar 31, 2016
  1. @znz

    Fix a typo

    znz committed Mar 31, 2016
Commits on Mar 23, 2016
  1. @schneems

    bump to v146

    schneems committed Mar 23, 2016
  2. @schneems

    v146

    schneems committed Mar 23, 2016
Commits on Mar 18, 2016
  1. @schneems
Commits on Mar 17, 2016
  1. @schneems
  2. @schneems

    Fix raise on fail behavior

    schneems committed Mar 17, 2016
  3. @schneems

    Merge pull request #474 from heroku/schneems/performance-m-web_concur…

    …rency
    
    Performance-M  WEB_CONCURRENCY data
    schneems committed Mar 17, 2016
  4. @schneems

    Merge pull request #471 from heroku/schneems/warn-bundle-config

    Warn on .bundle/config file presence
    schneems committed Mar 17, 2016
Commits on Mar 11, 2016
  1. @schneems
  2. @schneems

    Merge pull request #473 from heroku/mattgraham-patch-1

    Update buildpack image to align with marketing
    schneems committed Mar 11, 2016
  3. @mattgraham

    Update README.md

    mattgraham committed Mar 11, 2016
Commits on Mar 10, 2016
  1. @schneems

    Warn on .bundle/config file presence

    Discussion at bundler/bundler#4351 (comment)
    
    Also do not load the stored `.bundle` directory. We are passing in all the flags that we need to `bundle install`. If we also store a `.bundle/config` file between builds than if you accidentally include `.bundle/config` file in your repo with a `BUNDLE_WITHOUT` value ,it will get persisted to the `.bundle/config` file to the cache and deleting the file from your repo does not get rid of the problem. The config is un-needed, since we're using all flags all the time, let's not rely on the `.bundle/config` file.
    schneems committed Mar 10, 2016
Commits on Mar 8, 2016
  1. @schneems

    [ci skip] v145

    schneems committed Mar 8, 2016
Commits on Mar 4, 2016
  1. @schneems

    Merge pull request #462 from heroku/schneems/fail-on-assets-rails

    [close #358] Fail on Rake detect for Rails Apps
    schneems committed Mar 4, 2016
Commits on Mar 3, 2016
  1. @schneems

    Update ruby.rb

    schneems committed Mar 3, 2016
  2. @schneems

    Merge pull request #468 from heroku/schneems/fix-bundler-warning-more

    Use correct metadata directory
    schneems committed Mar 3, 2016
  3. @schneems
  4. @schneems

    [close #358] Fail on Rake detect for Rails Apps

    Rails applications should be reasonably expected to run `rake` when a Rakefile is present.
    
    Right now this is the behavior we support heroku/heroku-buildpack-ruby#358 (comment). While it might be reasonable that a Ruby application might not expect us to run Rake, a Rails app with the asset pipeline, while they might not be using the asset pipeline, should be able to run the `rake` command when a Rakefile is present.
    schneems committed Feb 29, 2016
  5. @schneems

    Update ruby.rb

    schneems committed Mar 3, 2016
  6. @schneems

    Update ruby.rb

    schneems committed Mar 3, 2016
  7. @schneems

    Merge pull request #465 from heroku/schneems/better-bundler-error2

    Alternative to #463 always warn on Bundler change
    schneems committed Mar 3, 2016
  8. @schneems

    Merge pull request #467 from heroku/schneems/fix-default-env-vars

    Fix config vars for existing apps
    schneems committed Mar 3, 2016