If two projects get updated at around the same time Goldberg should be able to run the projects in parallel (or even the parallel steps defined in #65). By default the maximum number of parallel builds should be based on the number of cores, but should be overrideable.

Provide a build status badge which can be embedded in READMEs etc. See the README of VCR for an example.

Can whoever that has access add the rspec-core project, running on JRuby 1.6.0 to http://goldberg.c42.in ?
The github url is https://github.com/rspec/rspec-core

See any of the project pages at http://golberg.c42.in

We need to be able to add projects through subversion. The default should remain git. You configure svn by passing a command line arg (say, --svn) or if the url is svn://...

I would expect that goldberg would do a "bundle install" after pulling the latest from git.

_rvm_reset:14: no matches found: /Users/tsedano/.rvm/bin/default*
_rvm_reset:14: no matches found: /Users/tsedano/.rvm/bin/passenger*
_rvm_reset:14: no matches found: /Users/tsedano/.rvm/bin/editor*
�[31mCould not find activesupport-2.3.4 in any of the sources�[0m
�[31mCould not find activesupport-2.3.4 in any of the sources�[0m

If I do a manual build install in the ~..goldberg/projects/PROJECT/code this goes away.

I should be able to configure the setting of environment variables for the build process as part of the project config. This would allow me to, for example, set SPEC="--profile" on my build so that rspec prints build stats.

In the even of a build hanging, Goldberg should kill it after a configurable period of time and mark the build as failed, clearly stating that it failed because it hung.

I should also be able to define steps in the build where one step (or multiple if #65 has already been implemented) has to be run before one (or multiple) other steps. I should be able to specify different configurations for each step.

I don't have RVM installed. When I start the poller, I get the following:

bash: -c: line 0: syntax error near unexpected token `;'
bash: -c: line 0: `( ; cd /root/.goldberg/projects/remindee/code;  (bundle check --no-color || bundle install --no-color) && rake default) 1>>/root/.goldberg/projects/remindee/builds/2/build_log 2>>/root/.goldberg/projects/remindee/builds/2/build_log'

I've only included the relevant part.

If I'm right the first part of this line returns nil: https://github.com/c42/goldberg/blob/master/app/models/build.rb#L58

I think we need to check if it returns nil and build an appropriate command string. What do you think?

I ran into an error trying to force a build on a new project I created. I was able to add the project using bin/goldberg add. After adding the project I launched the app with bin/goldberg start 3001. When I click force build on the project, nothing appears to happen (not sure if that is normal behavior), then when I click the project to see its builds, I get the standard Rails "We're sorry but something went wrong" error. Checking production.log yields the following error:

Started GET "/projects/amigo-master" for 172.16.0.70 at 2011-06-02 10:31:03 -0600
  Processing by ProjectsController#show as HTML
  Parameters: {"project_name"=>"amigo-master"}
Rendered projects/show.html.haml within layouts/application (1.8ms)
Completed   in 4ms

ActionView::Template::Error (undefined method `empty?' for nil:NilClass):
    22:       = @project.name
    23:       %em.project_ruby (#{@project.ruby})
    24:       %strong.project_status= "#{build_status(@build)}"
    25:     - unless @build.artefacts.empty?
    26:       %ul
    27:         Build artefacts
    28:         - @build.artefacts.each do |artefact|
  app/views/projects/show.html.haml:25:in `_app_views_projects_show_html_haml__3906049776127963241_30547500_2133775230302929165'

What is the best way to get around this?

Currently, you can only run the default rake target, but you should be able to run a different target, e.g. something else which runs both "db:migrate" & "spec".

Hi,

When trying to add a github project to Goldberg:

[deploy@li269-181 goldberg]$ bin/goldberg add [email protected]:tfw/ADA-CMS.git ada-cms          
executing /usr/bin/env bash -c "git clone --depth 1 [email protected]:tfw/ADA-CMS.git /home/deploy/.goldberg/projects/ada-cms/code --branch master"
error: unknown option `branch'
usage: git-clone [options] [--] <repo> [<dir>]

    -n, --no-checkout     don't create a checkout
    --bare                create a bare repository
    --naked               create a bare repository
    -l, --local           to clone from a local repository
    --no-hardlinks        don't use local hardlinks, always copy
    -s, --shared          setup as a shared repository
    --template ...        path to the template directory
    -q, --quiet           be quiet
    --reference ...       reference repository
    -o, --origin ...      use <name> instead of 'origin' to track upstream
    -u, --upload-pack ...
                          path to git-upload-pack on the remote
    --depth ...           create a shallow clone of that depth
    --use-separate-remote
                          compatibility, do not use
    --no-separate-remote  compatibility, do not use

ada-cms successfully added.
[deploy@li269-181 goldberg]$ bin/goldberg list
  Project Load (0.8ms)  SELECT "projects".* FROM "projects"

So, the project isn't added successfully - the branch arg isn't recognised - but the response message is that it's successful.

Adding with the -b arg results in the same.

[deploy@li269-181 goldberg]$ bin/goldberg add [email protected]:tfw/ADA-CMS.git ada-cms -b master
executing /usr/bin/env bash -c "git clone --depth 1 [email protected]:tfw/ADA-CMS.git /home/deploy/.goldberg/projects/ada-cms/code --branch master"
error: unknown option `branch'

See http://goldberg.c42.in/projects/wrest

• Move the list to the left column so I don't have to scroll to the bottom of a long build log
• Colour successes and failures differently - not necessarily the entire line but some part of it
• Link the SHA of the build to the repository browser if one exists - especially useful for Github projects

Repeated queries from CCMenu and other build notifiers shouldn't slow the system to a crawl.

This is a feature request

I'd like to configure goldberg to use rcov on my specs.

Running http://rubygems.org/gems/metrical or metrics_fu is a bonus.

Each build log should be prefixed with the following information:

• The equivalent of ruby -v
• The value of the RUBY_ENGINE constant, if it's available

Really enjoying this terrific build server yall have been kind enough to release.

Following the setup instructions in the README, I was able to get the server up and working by using rails s to boot Goldberg. If I boot it using bin/goldberg start, pages either never load, or errors are thrown (undefined method clear_active_connections! and an error related to page_cache).

I've tried editing the start script to explicitly use the mongrel server, which is what rails s would use, but it hasn't made any difference. Additionally, the server will sometimes flat-out fail when trying to boot, which results in a goldberg.pid file which must be removed by hand as the stop script doesn't delete it if the process is not found.

Thanks!

OS: Snow Leopard
Using RVM, ruby-1.9.2-p180

Have a page which can be exposed (without authentication) for public consumption, possibly without check in comments or committers names, just project name & a green or red status.

I checked the source out, did a bundle install. This failed.

gogo@rnjn ~goldberg(master) >bundle install
Fetching source index for http://rubygems.org/
Using rake (0.9.0)
Using abstract (1.0.0)
Using activesupport (3.0.6)
Using builder (2.1.2)
Using i18n (0.5.0)
Using activemodel (3.0.6)
Using erubis (2.6.6)
Using rack (1.2.3)
Using rack-mount (0.6.14)
Using rack-test (0.5.7)
Using tzinfo (0.3.27)
Using actionpack (3.0.6)
Using mime-types (1.16)
Using polyglot (0.3.1)
Using treetop (1.4.9)
Using mail (2.2.19)
Using actionmailer (3.0.6)
System.java:-2:in arraycopy': java.lang.ArrayIndexOutOfBoundsException from DefaultResolver.java:111:inmakeTime'
from DefaultResolver.java:277:in create' from DefaultResolver.java:317:inhandleScalar'

I should be able to specify (in the project config?) that the build should use a certain branch

See https://github.com/rubyspec/rubyspec/issues/2

Currently, a forced build doesn't do an update before running the build.

If you add a new project and goldberg starts a build, we get a missing image warning on the main page.

ActionController::RoutingError (No route matches "/images/not_available.png"):

The code that generates the request is here:

def project_status_image(status)
image_tag (status == 'building' ? "#{status}.gif" : "#{status}.png"), :alt => status, :title => status
end

I have a screenshot if its helpful.

(There's a side issue on why I'm getting this status at all. That may be worth discussing.)

Check if there exists a Gemfile (and possibly a versioned Gemfile.lock) and if there are changes bundle install before running the build.

Splitting off the third point of #16 raised by @kaiwren.

I'm looking at my build_log file and here is what I see....

_rvm_reset:14: no matches found: /Users/tsedano/.rvm/bin/default*
_rvm_reset:14: no matches found: /Users/tsedano/.rvm/bin/passenger*
_rvm_reset:14: no matches found: /Users/tsedano/.rvm/bin/editor*
zsh:1: command not found: bundle
zsh:1: command not found: bundle

a) my ~/.rvm/bin directory contains my languages (I don't see passenger, etc in there)
ruby-1.9.2-p180
ruby-1.9.2-p180@goldberg
ruby-1.8.7-p302@rails3
...

b) I'm not sure why zsh is failing to find bundler. Maybe we need to setup zsh to understand about rvm and which ruby to use? Normally I default my entire system to 1.8.7 and I have specific projects use 1.9.2 -- today, I turned off my default rvm and zsh couldn't find bundler.

Any thoughts?

Goldberg assumes there is an existing build for every project, and so would be failing for a new project till at least one build is started.

Runtimes that need to be added to http://goldberg.c42.in :

• Rubinius 1.2
• Ruby 1.8.6
• Ruby 1.8.7
• Ruby 1.9.2
• Jruby 1.6.0

I should be able to define a project to run builds of multiple configurations, but still have one status which is an && of the statuses of all the builds. I don't care about the order of the builds.

Goldberg seems to be deleting the Gemfile.lock even if it is versioned. https://gist.github.com/fa4eb6cc16d0ced1c0c2

Provide a mechanism (say, a pre-determined folder) that a build must do to expose build artefacts to the end user.

http://goldberg.ambekallu.com/projects/wrest/builds/999

The builds page currently looks like wrest (1.9.2) or rspec-mocks (jruby). The ruby version information should be expanded to include

• the runtime (MRI/YARV/JRUBY/RBX/MacRuby)
• the version (1.8.6/1.8.7/1.9.2)
• the patch number (1.9.2-p180)

... possibly through the use of a goldberg_settings.rb or some such that gets checked in.

Goldberg is working nicely on my development machine. I'm now installing it on my CI machine. (I stole my database.yml file from a cruise control project.) It only had the test database environment defined. It appears that Goldberg also wants the development database configured. I'm guessing we need a RAILS_ENV="test" on one of the rake commands to fix this.

Build successes/failures should be postable to IRC. For examples, look at #jruby's logs.

Create a CHANGELOG file under root so that supported features/fixed issues are clearly tracked in one place

I am encountering a simple issue where rake is not run with bundle exec with ruins the purpose of using bundler in the first place. I will attempt to find where the command is run and supply a patch.

The Gemfile's dependencies are satisfied
rake aborted!
You have already activated rspec-core 2.5.2, but your Gemfile requires rspec-core 2.4.0. Consider using bundle exec.
/home/ci/.goldberg/projects/amigo-master/code/Rakefile:4:in `<top (required)>'
(See full trace by running task with --trace)
(in /home/ci/.goldberg/projects/amigo-master/code)

Although I like how the build timestamps are shown in UTC, I'd like to be able to configure goldberg to show the time in my local time zone. I was tempted to add in this code, but I wasn't sure how the group felt about this item.

I have tried various incantations of the goldberg_config.rb file, but in every case, when I specify any commands to override (such as ruby= or environment_variables= or even command=), the build will fail without producing any output. However, if I leave the file in place and remove any assignments from the configure block, I get normal build output.

Here are a few sample configs I tried that produced no output on failed builds:

Project.configure do |c|
  c.ruby = '1.9.2-p0'
  c.environment_variables = {'RACK_ENV' => 'test', 'CHARSET' => 'latin1'}
  c.rake_task = 'ci'
end
Project.configure do |c|
  c.ruby = '1.9.2-p0'
  c.environment_variables = {'RACK_ENV' => 'test', 'CHARSET' => 'latin1'}
  c.command = 'bundle exec rake ci'
end
Project.configure do |c|
  c.ruby = '1.9.2-p0'
  c.command = 'bundle exec rake ci'
end
Project.configure do |c|
  c.command = 'bundle exec rake ci'
end

If the file is empty or non-existant, the build produces output normally. Also of note, the poller log produces only a single line of output on failed builds: "Build Failed". No warnings or stack traces are output.

This is a feature request

I'd like to configure goldberg.yml to use gmail to send out the "build failed" or "build succeeded" email. I would expect an email every time the build failed and one email when it started to succeed again.

bin/goldberg add xyz xyz

will create an entry in Projects even though the clone was not successful. The web UI also fails to load and silently fails after this.