ruby-git / ruby-git Goto Github PK

I've just tried this gem for the first time, and have not had any luck getting it to load. To be fair, I have installed ruby 2 on my Mac OS X Mountain Lion system using homebrew. But this doesn't seem to have caused me any similar problems with the many other rubygems I'm using.

Here I reproduce the problem with an interactive interpreter, using the default ruby that ships with Mac OS X 10.8:

aoide:kaelin.github.io kaelin$ env PATH=/usr/bin:/bin:/usr/sbin:/sbin:/usr/local/bin:/opt/X11/bin ruby --version
ruby 1.8.7 (2012-02-08 patchlevel 358) [universal-darwin12.0]
aoide:kaelin.github.io kaelin$ env PATH=/usr/bin:/bin:/usr/sbin:/sbin:/usr/local/bin:/opt/X11/bin ruby
require 'rubygems'
require 'git'

/Library/Ruby/Site/1.8/rubygems/core_ext/kernel_require.rb:45:in `gem_original_require': no such file to load -- git (LoadError)
    from /Library/Ruby/Site/1.8/rubygems/core_ext/kernel_require.rb:45:in `require'
    from -:2

Here is the stack :

/usr/lib/ruby/1.8/git/lib.rb:148:in process_commit_data': undefined method<<' for nil:NilClass (NoMethodError)
from /usr/lib/ruby/1.8/git/lib.rb:135:in each' from /usr/lib/ruby/1.8/git/lib.rb:135:inprocess_commit_data'
from /usr/lib/ruby/1.8/git/lib.rb:123:in commit_data' from /usr/lib/ruby/1.8/git/object.rb:230:incheck_commit'
from /usr/lib/ruby/1.8/git/object.rb:197:in committer' from /usr/lib/ruby/1.8/git/object.rb:202:incommitter_date'
from /home/djdeath/bin/confront_patch_list.rb:32
from /home/djdeath/bin/confront_patch_list.rb:27:in `each_key'
from /home/djdeath/bin/confront_patch_list.rb:27

It happens when I call the committer_date method with something like this :

repo = Git::Base.open("#{ENV['PWD']}")
ob = Git::Object::Commit.new(repo, "whatevercommithash")
puts ob.committer_date

The problem is in the process_commit_data method in lib.rb because this method expects the commit's metadatas to be separated from the commit's comment by an empty line. But it does not expect the commit to contain a empty line too.

Here is what the commit triggering this crash looks like :

$ git cat-file commit a693b0cdba94f60f7ed43754d2c34151cdd11da5
tree 99cbf46748dfa4f35acc1d63ae51817f3e17ec9d
parent 484ab62c5ee805c2bdc405a85a4e64da2722690f
author Mauro Carvalho Chehab [email protected] 1228161854 -0200
committer Mauro Carvalho Chehab [email protected] 1228161854 -0200

em28xx: remove backward compat macro added on a previous fix

commit 50f3beb50abe0cc0228363af804e50e710b3e5b0 fixed em28xx-alsa
locking schema. However, a backport macro was kept.

This patch removes the macro, since it is not needed for the module
compilation against upstream.

Signed-off-by: Mauro Carvalho Chehab [email protected]

So the fix is pretty easy, on lib.rb line 137 you have :

    if line == ''

Replace it by :

    if line == '' and not in_message                                       

Regards,

Lionel Landwerlin

g = Git.clone(URI, :name => 'name', :path => '/tmp/checkout') 
TypeError: can't convert Hash into String                                                                                                  

should be
g = Git.clone(URI, 'name', :path => '/tmp/checkout')

Looks like the way that the ArgumentError exception is rised in not compatible with JRuby.

===============================================================================

Failure:

test_initialize_with_bad_path_and_check_path(TestGitPath)

/home/travis/build/robertodecurnex/ruby-git/tests/units/test_git_path.rb:18:in `test_initialize_with_bad_path_and_check_path'

15: end

16:

17: def test_initialize_with_bad_path_and_check_path

=> 18: assert_raises ArgumentError do

19: Git::Path.new('/this path does not exist', true)

20: end

21: end

<ArgumentError> expected but was

<TypeError(<backtrace must be Array of String>)>

#58 Adds clone --recursive support.

We do not support submodules yet so we should add tests over #58 right after we add submodules support.

(Submodules support has started at #19)

I get following errors on linux with git version 1.7.4.1 and ruby-git gem version 1.2.5 when color ui is set to always.

require 'git'; git = Git.open('.'); git.log(1).inspect
NoMethodError: undefined method `[]=' for nil:NilClass
    from /home/konrad/.rvm/gems/ruby-1.9.2-p180@example_gemset/gems/git-1.2.5/lib/git/lib.rb:153:in `block in process_commit_data'
    from /home/konrad/.rvm/gems/ruby-1.9.2-p180@example_gemset/gems/git-1.2.5/lib/git/lib.rb:135:in `each'
    from /home/konrad/.rvm/gems/ruby-1.9.2-p180@example_gemset/gems/git-1.2.5/lib/git/lib.rb:135:in `process_commit_data'
    from /home/konrad/.rvm/gems/ruby-1.9.2-p180@example_gemset/gems/git-1.2.5/lib/git/lib.rb:93:in `full_log_commits'
    from /home/konrad/.rvm/gems/ruby-1.9.2-p180@example_gemset/gems/git-1.2.5/lib/git/log.rb:108:in `run_log'
    from /home/konrad/.rvm/gems/ruby-1.9.2-p180@example_gemset/gems/git-1.2.5/lib/git/log.rb:101:in `check_log'
    from /home/konrad/.rvm/gems/ruby-1.9.2-p180@example_gemset/gems/git-1.2.5/lib/git/log.rb:84:in `each'
    from /home/konrad/.rvm/gems/ruby-1.9.2-p180@example_gemset/gems/git-1.2.5/lib/git/log.rb:72:in `map'
    from /home/konrad/.rvm/gems/ruby-1.9.2-p180@example_gemset/gems/git-1.2.5/lib/git/log.rb:72:in `to_s'
    from (irb):4:in `inspect'
    from (irb):4
    from /home/konrad/.rvm/rubies/ruby-1.9.2-p180/bin/irb:16:in `<main>'

Solution is to add --no-color option to git commands like git log or git diff, so that color tags like '\e[43m' are not included in data being processed.

Hi!

I'm running version 1.2.5 of the gem and 1.9.2 (p290) of ruby.
The following code crashes if the log contains a foreign character in the git log:

require 'git'

# Init git and get log
g = Git.init
log = g.log

changelog = "My changes: \n"

# Now create changelog (and crash)
log.each {|l| 
   changelog += "- " + l.message + "\n")
}

Stack trace:

[...]/gems/git-1.2.5/lib/git/lib.rb:672:in `split': invalid byte sequence in US-ASCII (ArgumentError)
   from [...]/gems/git-1.2.5/lib/git/lib.rb:672:in `command_lines'
   from [...]/gems/git-1.2.5/lib/git/lib.rb:92:in `full_log_commits'
   from [...]/gems/git-1.2.5/lib/git/log.rb:108:in `run_log'
   from [...]/gems/git-1.2.5/lib/git/log.rb:101:in `check_log'
   from [...]/gems/git-1.2.5/lib/git/log.rb:84:in `each'
   from gittest:10:in `<main>'

Any ideas?

I don't know if it is a bug, but "git status" reports nothing to commit whereas the Git::Status contains a list of untracked files. They are untracked because the are in a repository in repository (no the best idea, I know..).

How can I ignore the nested repos?

g = Git.open "repo"
g.pull
# "Already up-to-date."
g.lib.send(:command, 'pull')
# "Updating 6d57a5a..0bbe197\nFast-forward\n file.rb |    6 +-----\n 1 files changed, 1 insertions(+), 5 deletions(-)"

It looks like internally Git::Base#pull doesn't use git pull at all, but fetches and merges. Why does it not use git pull as expected? Currently it just reports it's already up to date.

r.remove
Git::GitExecuteError: git remote "rm" "--" "staging"  2>&1:usage: git remote rm <name>

If I do the following I can get it to work

git.lib.send(:command, 'remote', ['rm', 'staging'])

It seems git doesn't like the '--'

I am on git 1.7.7.3

I will try updating to the latest git and trying again.

When calling git.status files and directories starting with dot (.) are ignored. git status, on the other hand, does list such files and directories.

Seems like the repo path is not random enough so some times we end up trying to create/clone a new repository into an already taken path.

==============================================================================

Error: test_git_init_remote_git(TestInit)

Errno::EEXIST: File exists - /tmp/git_test1365775072186

/home/travis/.rvm/rubies/ruby-1.8.7-p371/lib/ruby/1.8/fileutils.rb:243:in `mkdir'

/home/travis/.rvm/rubies/ruby-1.8.7-p371/lib/ruby/1.8/fileutils.rb:243:in `fu_mkdir'

/home/travis/.rvm/rubies/ruby-1.8.7-p371/lib/ruby/1.8/fileutils.rb:172:in `mkdir'

/home/travis/.rvm/rubies/ruby-1.8.7-p371/lib/ruby/1.8/fileutils.rb:171:in `each'

/home/travis/.rvm/rubies/ruby-1.8.7-p371/lib/ruby/1.8/fileutils.rb:171:in `mkdir'

/home/travis/build/robertodecurnex/ruby-git/tests/units/../test_helper.rb:48:in `in_temp_dir'

/home/travis/build/robertodecurnex/ruby-git/tests/units/test_init.rb:42:in `test_git_init_remote_git'

39: end

40:

41: def test_git_init_remote_git

=> 42: in_temp_dir do |dir|

43: assert(!File.exists?(File.join(dir, 'config')))

44:

45: in_temp_dir do |path|

==============================================================================

I'm unclear what the equivalent command to

git log --reverse "filename"

would be, and how to get the timestamp of the first commit to that file (as would be returned in that log). e.g. without the reverse, I see I can get the most recent commit timestamp with

g = Git.open("path/to/repo")
g.log(1).object("filename").first.date

(Or perhaps there is a better way to get the first/oldest timestamp of a commit to a file using the gem?)

When having done a clone of a repository as bare with a depth of 1, in order to access a file ruby-git should implement a show method that will return the file content for given ref and file path.

I'm using ruby-1.9.2-p136 and irb:

g = Git.clone("https://github.com/opsb/ruby-git.git", :name => 'name', :path => '/tmp/checkout1')
TypeError: can't convert Hash into String
from /Users/guillaumeh/.rvm/gems/ruby-1.9.2-p136/gems/git-1.2.5/lib/git/base.rb:56:in `join'
from /Users/guillaumeh/.rvm/gems/ruby-1.9.2-p136/gems/git-1.2.5/lib/git/base.rb:56:in `initialize'
from /Users/guillaumeh/.rvm/gems/ruby-1.9.2-p136/gems/git-1.2.5/lib/git/base.rb:51:in `new'
from /Users/guillaumeh/.rvm/gems/ruby-1.9.2-p136/gems/git-1.2.5/lib/git/base.rb:51:in `clone'
from /Users/guillaumeh/.rvm/gems/ruby-1.9.2-p136/gems/git-1.2.5/lib/git.rb:97:in `clone'
from (irb):7
from /Users/guillaumeh/.rvm/gems/ruby-1.9.2-p136/gems/railties-3.0.3/lib/rails/commands/console.rb:44:in `start'
from /Users/guillaumeh/.rvm/gems/ruby-1.9.2-p136/gems/railties-3.0.3/lib/rails/commands/console.rb:8:in `start'
from /Users/guillaumeh/.rvm/gems/ruby-1.9.2-p136/gems/railties-3.0.3/lib/rails/commands.rb:23:in `<top (required)>'
from ./script/rails:6:in `require'
from ./script/rails:6:in `<main>'

Hello,
First, I'm sorry if there's a way to do it but I've searched, tried and even asked but I couldn't find a way so I decided to submit this as a new issue.

Let's assume I'm cloning a remote repository into a bare repository on my local machine. I can't do "g.pull" as this is just a bare repository so I can only use "g.fetch" which gets the newest updates but using "g.log" doesn't give me latest commits. It instead gives commits at the time of cloning. If I try to merge, it says that merging needs a working tree so I can't find any way to get updates to a bare repository to be just as if the repository was cloned at that time.

Thanks.

It seems like I'm unable to create bare repos: https://github.com/schacon/ruby-git/blob/master/lib/git/lib.rb#L29

The clone function here honors the :bare option: https://github.com/schacon/ruby-git/blob/master/lib/git/lib.rb#L49

The escape mechanism for handling filenames with single quotes in them is repeating text, not escaping at all.

s = "it's"
# Original
escaped = s.to_s.gsub('\'', '\'\\\'\'') # => "it's's"
%Q{"#{escaped}"} #=> "\"it's's\""

# Patched
%Q{"#{s.to_s.gsub("'") { "\\'" }}"} # => "\"it\\'s\""

Looks like gsub placeholders are kicking in accidentally.

When performing a local git clone, the newly cloned git repository can use the objects stored in the original git repository by use of '.git/objects/info/alternatives' instead of copying objects. In cases where a large number of clones of the same git repository are being done, this can speed up clone operations by a significant amount and deduplicate git objects.

To add this functionality, the following changes would need to be made.

Git::Lib#clone would need to support a :reference or :shared option to use the alternates object directory on initial clone.

A new class like Git::Object::Alternate or Git::Alternate should be added to represent an entry in the alternates file. A member object should be able to generate a new Git::Base object for the referenced repository.

Git::Base should define a method like #alternates or something equivalent that would list the entries in .git/objects/info/alternates. It would return an array of zero or more Git::Alternate entries.

I would be very happy to help out with the implementation of this feature, and I would be fine to implement the entire thing myself. I'm opting to file this feature request so that I can figure out the best way to make this change before I start submitting pull requests.

As a reference, I'm currently implementing this functionality at https://github.com/adrienthebo/r10k/blob/master/lib/r10k/git/working_dir.rb#L79 by directly invoking git clone --reference, but I would rather move to ruby-git for this behavior instead of reinventing the wheel.

Thank you for your time!

@schacon maybe time for The Pull Request Hack? It's been a long while since I've been hacking on this myself, so time to open the flood gates on collaboratorness?

On JRuby calling #add inside the #with_temp_index block changes the tree name every time, even if there are no changes between the first and second call.

I'm attempting to check for git changes in a current directory after I re-write a file. For some reason @git.status.changed always shows that the file has been changed, even though command line git status shows no diff, as it should. Is there a way around this. I'm attempting to decide whether to update a field in the file if it has been changed before pushing to git.

    # Check the if the contents of a directory in git
    def dir_has_changes? dir
        intersection = dir.split('/') - @top_level.split('/')
        intersection = intersection.join('/')
        has_changes = false
        @git.status.changed.each do |file|
            if file[0] =~ /#{intersection}/
                has_changes = true
            end
        end
        return has_changes
    end

There's no way to send options such as --force to Git::Base#add.

Fixed with a patch: http://github.com/tadman/ruby-git/commit/de6ad7dfd1d936cb789f055993f67aa8af25b599

Given a commit like this:

commit 4431b6b764e80d1db038857dbb699948268d1f84
Author: Brian xxxxx [email protected]
Date: Mon Nov 28 13:14:57 2011 -0500

Adding a modified post-receive script for extensibility

The post-recieve script should allow for adding multiple
sub-scripts as necessary though iteritive invocation.

o = repo.object('4431b6b764e80d1db038857dbb699948268d1f84')
o.message
=> "Adding a modified post-receive script for extensibility"

The problem is in lib.rb#process_commit_data. The assumption in the current code is if line.empty? then this should mark the beginning or ending of processing of the commit message. I tried patching it, but even without my code changes tests are not passing on trunk. My changes make the tests break more, so it's not solid, but here's a start:

https://gist.github.com/1412804

/opt/local/lib/ruby1.9/gems/1.9.1/gems/schacon-git-1.2.2/lib/git/lib.rb:345:in chdir': wrong number of arguments (1 for 0) (ArgumentError) from /opt/local/lib/ruby1.9/gems/1.9.1/gems/schacon-git-1.2.2/lib/git/lib.rb:345:inconfig_list'
from /opt/local/lib/ruby1.9/gems/1.9.1/gems/schacon-git-1.2.2/lib/git/lib.rb:315:in config_remote' from /opt/local/lib/ruby1.9/gems/1.9.1/gems/schacon-git-1.2.2/lib/git/remote.rb:8:ininitialize'
from /opt/local/lib/ruby1.9/gems/1.9.1/gems/schacon-git-1.2.2/lib/git/base.rb:330:in new' from /opt/local/lib/ruby1.9/gems/1.9.1/gems/schacon-git-1.2.2/lib/git/base.rb:330:inblock in remotes'
from /opt/local/lib/ruby1.9/gems/1.9.1/gems/schacon-git-1.2.2/lib/git/base.rb:330:in map' from /opt/local/lib/ruby1.9/gems/1.9.1/gems/schacon-git-1.2.2/lib/git/base.rb:330:inremotes'

http://wiki.github.com/schacon/ruby-git indicates that grit is the library to use, and if that's the case, the readme on ruby-git should reflect this.

The native git command has the --follow option to see logs of files that are moved / renamed.

Could you please take a look and support an option for the .log operation, for example

$repo.object($object).log(10, true)
#10: number of log entriess
# true: continue listing the history of a file beyond renames

Thank you

The 1.2.5 gem does not contain the tests. It would be great to include them in the gem so that it is easy to run them locally.

Right now we use TempFile to create temp paths.

It was not working for JRUBY so a little hack was applied to go over it [https://github.com//issues/72].

Also, creating a file, to be deleted right after that just to get a random temp path does not seems right.

Example (with the JRUBY workaround):

if RUBY_PLATFORM == 'java'
  temp_path = "/tmp/temp-index-#{(0...15).map{ ('a'..'z').to_a[rand(26)] }.join}"
else
  tempfile = Tempfile.new('temp-index')
  temp_path = tempfile.path
  tempfile.close
  tempfile.unlink
end

Git:Log class has the first attribute, why it doesn't also have last attr ?

When name and e-mail are quoted in ~/.gitconfig, such as:
[user]
name = "Mister Dude"
email = "[email protected]"

the Rakefile that Jeweler generates for a new project double quotes author and email:
Jeweler::Tasks.new do |gem|
gem.name = "the-perfect-gem"
gem.summary = %Q{TODO}
gem.email = ""[email protected]""
gem.homepage = "http://github.com/foo/the-perfect-gem"
gem.authors = [""Mister Dude""]
end

This is because ruby-git parses ~/.gitconfig directly instead of using the system commands.

Hey,

im currently writing a script which is pushing git commits to the server..
Sometimes they are really big and need some minutes to transfer.. and there is no chance to get out how many percent of the transmission is already finished..

could you implement something like

Tread.new { g.push }

10.times do
g.status.push_delivered_in_percent
end
=> 10 %, 20 %,...

Thank you!
If you have any other suggestions, please feel free to tell me!

The current implementation of the methods in lib.rb accepts options, but only passes a select subset of those options through to #command. It you want to do something like
--amend a commit you cannot use the commit method, and have to send('command') with the appropriate options.

Can the structure be changed to pass through the options to the command method to support all git options for these commands?

Any reason you do not have ",opts = {}" as a "def push" parameter in base.rb?

I am trying to do git push -u and found only "tags".

Is there another way to do "-u" instead of changing the code?

Thanks.

Travis fails because there's no git user configured in the environment.

 Git::GitExecuteError: git commit "-m" "test" "-a" 2>&1:

*** Please tell me who you are.

Run

git config --global user.email "[email protected]"

git config --global user.name "Your Name"

to set your account's default identity.

Omit --global to set the identity only in this repository.

This was working before #66
The error cause is not obvious and its clearly related with the Ruby 2.0.0 internals.

===============================================================================

Error: test_conflicts(TestEachConflict)

Errno::ENOENT: No such file or directory - /tmp/YOUR-example.txt20130412-1113-1rren04

/home/travis/build/robertodecurnex/ruby-git/tests/units/test_each_conflict.rb:39:in `read'

/home/travis/build/robertodecurnex/ruby-git/tests/units/test_each_conflict.rb:39:in `block (3 levels) in test_conflicts'

/home/travis/build/robertodecurnex/ruby-git/lib/git/lib.rb:525:in `block in conflicts'

/home/travis/build/robertodecurnex/ruby-git/lib/git/lib.rb:519:in `each'

/home/travis/build/robertodecurnex/ruby-git/lib/git/lib.rb:519:in `conflicts'

/home/travis/build/robertodecurnex/ruby-git/lib/git/base.rb:319:in `each_conflict'

/home/travis/build/robertodecurnex/ruby-git/tests/units/test_each_conflict.rb:37:in `block (2 levels) in test_conflicts'

/home/travis/build/robertodecurnex/ruby-git/tests/units/test_each_conflict.rb:16:in `chdir'

/home/travis/build/robertodecurnex/ruby-git/tests/units/test_each_conflict.rb:16:in `block in test_conflicts'

/home/travis/build/robertodecurnex/ruby-git/tests/test_helper.rb:53:in `block in in_temp_dir'

/home/travis/build/robertodecurnex/ruby-git/tests/test_helper.rb:52:in `chdir'

/home/travis/build/robertodecurnex/ruby-git/tests/test_helper.rb:52:in `in_temp_dir'

/home/travis/build/robertodecurnex/ruby-git/tests/units/test_each_conflict.rb:14:in `test_conflicts'

11: end

12:

13: def test_conflicts

=> 14: in_temp_dir do |path|

15: g = Git.clone(@wbare, 'branch_merge_test')

16: Dir.chdir('branch_merge_test') do

17:

===============================================================================

From the documentation for git-clone(1):

--no-checkout
-n

No checkout of HEAD is performed after the clone is complete.

This feature is useful if you're updating a repository from some other source. In my case I have a script run by a cron job that grabs the latest copy of source from a central place that is out of my control, and then updates my local git repo by doing, basically:

git clone -n ...
git checkout .gitignore
extract_from_remote
git commit/push ...

I'm planning to port this script to ruby-git and this issue blocks.

SinceGit::Base#in_remote_branch? and Git::Base#is_local_branch? use the Git::Branch#name to find a branch it wont work for remote branches sharing the same #name as a local one.

As an example, you can not check if origin/master is remote or not.

I'm trying to build a tool that allows people to enter a SHA and see all refs that point to that SHA, but can't do it with ruby-git because the git log command does not support annotate.

I've been trying to find a memory leak in a rails app that I'm making. A particular request is increasing the apps memory usage with each request. I stripped everything out of the controller and view for that request and observed that the apps memory usage was now static when it was called. Then line by line I added the controller and view code back in.
The point at which the memory usage started increasing again was a call to a method which in turn called .status on a repo. With everything removed except the call to repo.status the memory usage was still increasing. Without out the call to repo.status the memory usage stayed constant.

I did a simple test in a fresh irb. Connected to a repo and called status and observed the consoles memory usage.

git = Git.open(<path_to_repo>)
git.status

each time I called git.status I could see the consoles mem usage increase by around 80kb. If I stopped calling git.status the mem usage stayed constant.
I found this to be the case on both Linux and Windows dev machines.

I'm sorry if I'm barking up the wrong tree with this, I hope I'm not wasting your time!

From inline notes,

http://github.com/schacon/ruby-git/commit/96bf1ac28377bdaa6b26fbaf160894c2e99c6cd1

This is a patch taken over from Grit, to correct an error in Ruby-Git.

http://github.com/riyad/grit/blob/master/lib/grit/status.rb#L219

Completes the list of ignored files by globbing the ignored directories. Normally git won't recurse into ignored directories. So we use Dir.glob instead.

foo = REPOSITORY.object('HEAD')

foo.sha => "b64400d7dc2dd23474120e12c7bca59b41be771a"

foo.date => Tue Jul 21 14:55:16 -0700 2009

foo.sha => "HEAD"

Maybe I'm missing something, but it seems like Git.clone should act like Git.open:

g = Git.clone (url, name, :log => Logger.new(STDOUT))

vs.

g = Git.open (working_dir, :log => Logger.new(STDOUT))

But it doesn't

The homepage URL of this project does not work.
It is given on this project's site as well as in the README file.

Hi,

How can I fetch all commits for all the branches of a given repository? Is this possible ?

Git::GitExecuteError: git diff "-p" "6ae73defe24307d1bca14a9e935cf370aeb016a4" "" 2>&1:fatal: ambiguous argument '': unknown revision or path not in the working tree. Use '--' to separate paths from revisions

I'm trying to use your ruby-git gem to write some helper scripts for my own use but
+g.pull(remote, branch) doesn't have the expected behavior.

I want to process a pull request, so I extract the git url and the branch name from
+the mail and call:

g.pull(url, branch)

the problem is both on fetch and merge.

If I pass a branch to be pulled, you should probably go and fetch only that branch,
+but currently you only fetch the current checked out HEAD (which is master most of
+the times).

Also, you always overwrite the cool git message "merge branch of into
+" with "origin pull" which isn't really useful.

Then you try to merge an unexistent branch. What you should, probably, do is something
+like:

def pull(remote = 'origin', branch = 'master', message = nil)
fetch(remote, branch)
merge('FETCH_HEAD', message)
end

Here's the output of my script:

I, [2010-04-27T09:58:39.566725 #3309] INFO -- : Starting Git
I, [2010-04-27T09:58:39.570332 #3309] INFO -- : git branch newbranch-1272284431 2>&1
I, [2010-04-27T09:58:40.233505 #3309] INFO -- : git checkout newbranch-1272284431 2>&1
I, [2010-04-27T09:58:40.630576 #3309] INFO -- : git reset --hard origin/master 2>&1
D, [2010-04-27T09:58:40.630677 #3309] DEBUG -- : HEAD is now at bf4f324 Merge branch 'testing' into master
I, [2010-04-27T09:58:40.945683 #3309] INFO -- : git fetch git://mydomain.com/tree.githttp://mydomain.com/tree.git 2>&1
D, [2010-04-27T09:58:40.945802 #3309] DEBUG -- : From git://mydomain.com/treehttp://mydomain.com/tree

• branch HEAD -> FETCH_HEAD
  I, [2010-04-27T09:58:40.968073 #3309] INFO -- : git merge -m 'origin pull' branch 2>&1
  D, [2010-04-27T09:58:40.968177 #3309] DEBUG -- : fatal: branch - not something we can merge
  /usr/lib/ruby/1.8/git/lib.rb:643:in command': git merge -m 'origin pull' branch +2>&1:fatal: branch - not something we can merge (Git::GitExecuteError) from /usr/lib/ruby/1.8/git/lib.rb:467:inmerge'
  from /usr/lib/ruby/1.8/git/base.rb:304:in merge' from /usr/lib/ruby/1.8/git/base.rb:317:inpull'
  from ./test.rb:37
  from ./test.rb:28:in `each_line'
  from ./test.rb:28

I changed the branch names and tree addresses, of course.

git diff --name-only --staged

Hi!

It happens from time to time that the status returned from the git gem is incorrect. I have no idea why it happens sometimes and not others. For example today:

$ ruby -rubygems -e 'require "git"; p Git.open(".").status.changed'
[["vendor/mead/etc/enidf", #<Git::Status::StatusFile:0x1021849f0 @mode_repo="100644", @base=#<Git::Base:0x1022e4f70 @logger=nil, @lib=#<Git::Lib:0x1022e4cc8 @logger=nil, @git_index_file="/Users/mal/project/.git/index", @path=nil, @git_dir="/Users/mal/project/.git", @git_work_dir="/Users/mal/project">, @index=#<Git::Index:0x1022e4d90 @path="/Users/mal/project/.git/index">, @working_directory=#<Git::WorkingDirectory:0x1022e4e08 @path="/Users/mal/project">, @repository=#<Git::Repository:0x1022e4de0 @path="/Users/mal/project/.git">>, @untracked=nil, @type="M", @path="vendor/mead/etc/enidf", @mode_index="100644", @sha_repo="b63bbe70c7afcb44ddacc6de272351febd171207", @stage="0", @sha_index="0000000000000000000000000000000000000000">]]
$ git status
# On branch master
# Your branch is ahead of 'origin/master' by 2 commits.
#
nothing to commit (working directory clean)


$ ruby -v
ruby 1.8.7 (2009-06-08 patchlevel 173) [universal-darwin10.0]
$ gem which git
/Library/Ruby/Gems/1.8/gems/git-1.2.5/lib/git.rb
$ git --version
git version 1.7.0.3

A git reset HEAD --hard won't fix it, so I delete the files first, then check them out and now the status is ok (ie. [])