Commits on Apr 25, 2015
Commits on Apr 2, 2014
  1. Merge pull request #31 from kiwiroy/master

    Update design.pod
    trivial fix to pod to silence pod errors
    bucciarati committed Apr 2, 2014
  2. Update design.pod

    trivial fix to pod to silence pod errors
    kiwiroy committed Apr 2, 2014
Commits on Oct 24, 2013
  1. Merge pull request #29 from guest20/master

    fixed a bug where %TAG isn't expanded
    dgryski committed Oct 24, 2013
Commits on Sep 3, 2013
  1. Minor correction to language

    Changing the order of words to be correct English.
    [Ævar: Re-do this commit to also patch bin/git-deploy which
    is generated from]
    mfjordvald committed with avar Sep 3, 2013
  2. Merge pull request #28 from mfjordvald/patch-1

    Minor correction to language
    dgryski committed Sep 3, 2013
  3. Minor correction to language

    Changing the order of words to be correct English.
    mfjordvald committed Sep 3, 2013
Commits on Jul 29, 2013
  1. README.pod: bump from git-deploy

    In particular, make the licensing information more obvious.
    Updated with: grep -A 9001 ^=head1 git-deploy >README.pod
    dgryski committed Jul 29, 2013
Commits on Mar 18, 2013
  1. Add a "push-hotfix" action to automate hotfix recipe

    See Issue #25 - TODO Eliminate busywork from hotfix process.
    This is an untested initial attempt at a recipe for this process. It is
    interesting as we dont have an easy way currently to the lock a directory
    without modifying the logfile. We solve this by holding a lock on the
    logfile for the duration of the action.
    IOW, this includes code to support a "logless lock". (Quite simple logic
    demerphq committed Jan 6, 2013
Commits on Feb 5, 2013
  1. git-deploy: document the post-finish hook

    In particular suggest that E-Mails be sent out there instead of in
    post-sync due to #27
    avar committed Feb 5, 2013
  2. Add post-tag again with better behavior and documentation

    Re-add the post-tag hook Hans was trying to add in
    5c4b77f, but actually run it every
    time we create a tag, and set "rollout_tag" in the argument to the
    hook, also document it in the POD.
    avar committed Feb 5, 2013
  3. Remove the _make_dated_tag function

    Finish what I started in 9a84cf6 and just remove the _make_dated_tag()
    function. It was only called from make_and_push_dated_tag().
    avar committed Feb 5, 2013
  4. git-deploy: don't redundantly push_tag() after make_and_push_dated_tag()

    The make_and_push_dated_tag() call is going to push_tag(), no need to
    do it again.
    avar committed Feb 5, 2013
  5. Revert "Added a post-tag hook."

    This reverts commit 5c4b77f. Going to
    re-do this as something that runs the post-tag hook on
    make_and_push_dated_tag() as discussed with Hans.
    avar committed Feb 5, 2013
  6. make_and_push_dated_tag: just push the tag we created, not all tags

    In theory we should now be pushing all the tags we create, so why do
    we even need this:
        # and push any tags that might have been created
        push_tags( $remote_site );
    But just push the tag we created here, if there's any leftover ones
    for some reason they'll be picked up by the above code.
    avar committed Feb 5, 2013
  7. git-deploy: fix compilation error in f348aa2

    The f348aa2 commit didn't compile because we weren't passing
    $remote_site around.
    Maybe we should just move the do_interactive_revert function around to
    avoid this:
        bin/git-deploy:161:sub do_interactive_revert {
        bin/git-deploy:768:                do_interactive_revert( $remote_site, $list_count, $prefix, $action, $ignore_older_than, $include_branches,
    avar committed Feb 5, 2013
  8. Make the make_dated_tag function private

    This function hasn't been used by git-deploy since f348aa2, add a _
    prefix to it to indicate we only use it in Git::Deploy itself.
    avar committed Feb 5, 2013
  9. Merge branch 'github/issue_24/push_tags'. Resolves Issue #24

    Merge Yves's commit for pushing tags right after they're created. See
    avar committed Feb 5, 2013
  10. Added post-finish hook.

    Hans Tool committed Feb 5, 2013
Commits on Feb 4, 2013
  1. Added a post-tag hook.

    Hans Tool committed Feb 4, 2013
Commits on Jan 28, 2013
  1. Merge pull request #26 from burak/master

    Prevent recursion in _die()
    Looks good to me, tested it on the csapp box with a syntax error in git-deploy.conf.
    avar committed Jan 28, 2013
  2. Safegueard against recursion.__log() calls _get_config()

    which in turn might call _die() to handle errors.
    This will result with some meaningless errors as the
    actual error  will not reach the screen.
    Burak Gursoy committed Jan 28, 2013
Commits on Jan 22, 2013
  1. Merge remote-tracking branch 'github/master'

    Bring in upstream changes from GitHub which was ahead of us.
    avar committed Jan 22, 2013
  2. Revert "Changed output."

    This reverts commit 964c1d6. This was
    already solved upstream by
    avar committed Jan 22, 2013
  3. Changed output.

    Hans Tool committed Jan 22, 2013
Commits on Jan 6, 2013
  1. Push tags right after we create them.

    This is tracked as Issue #24. See also
    6c2947b for an alternate
    demerphq committed Jan 6, 2013
  2. Untangle issues with END blocks, and exit codes (introduce fatal_exit)

    When $? is set inside of an END block it changes the exit code returned
    to the OS. (See the docs for $? in perlvar.)
    We were calling routines that would execute system/backticks inside of
    an END block in Git::Deploy::Timing, which would clobber the exit code.
    Various patches ensued to try to work around this fact, but ultimately
    confused the matter more. This patch effectively reverts those patches,
    and makes it so we get configuration for Git::Deploy::Timing at startup
    and not shutdown, thus leaving the exit code alone during teardown.
    This patch also introduces a new sub exported by Git::Deploy::Say called
    fatal_exit($code,@msg) which is intended for top level routines to use for
    fatal exits which should return a specific exit code. Internal routines
    should continue to use _die() as it is trapable wheras fatal_exit() is
    not. In the long run we need to rething this separation of
    demerphq committed Jan 6, 2013
  3. Invoke a stable 'git status' command to check for cleanliness.

    Git 1.8.0 changed the format of 'git status' from
        ... (working directory clean)
        ..., working directory clean
    This kind of harmless modification of prose output happens pretty regularly,
    and relying on the prose parts to determine the result of a Git command is
    pretty fragile.
    Fortunately, Git folks are aware of this, and provide stable interfaces too.
    For 'git status', the programmatic interface is 'git status --porcelain',
    which has a stable (and much simpler) output format, with one line per
    non-clean file. Empty output indicates a clean tree.
    I haven't spent a huge amount of effort finding other places this might bite;
    with this change git-deploy works (to the best of my testing, which isn't much)
    on Git 1.8.0.
    Note we cannot just use `git status --porcelain`'s output to report to
    the user, as it would produce some ugly output:
        :)owen@Dashie:example (master)$ git deploy start
        # FATAL: working directory not clean. git status reports:
        # FATAL: ?? .deploy
        # FATAL: ?? foo
        Died at /Users/owen/Development/git-deploy/git-deploy line 45.
    Se we use `git status --porcelain` for checking the cleanliness of the work tree,
    but `check_if_working_dir_is_clean` once again returns non-`--porcelain` output,
    returning the UI to a sensible state:
        :(owen@Dashie:example (master)$ git deploy start
        # FATAL: working directory not clean. git status reports:
        # FATAL: On branch master
        # FATAL: Untracked files:
        # FATAL: (use "git add <file>..." to include in what will be committed)
        # FATAL: .deploy
        # FATAL: foo
        # FATAL: nothing added to commit but untracked files present (use "git add" to track)
        Died at /Users/owen/Development/git-deploy/git-deploy line 45.
    Putting the second call inside of `check_if_working_dir_is_clean` keeps the changes
    pretty localized, even though this sub only has one caller.
    Note that there is not much point in "optimizing" this code, for
    instance by rolling our own "pretty" status mode: the file system will
    be hot after the initial status, and the second should be quite fast as
    a consequence.
    Committers Note: This is a rebased/squash of two commits, with edits to
    the message. Also, the code this patch refers to predates the options this patch
    uses to fix things, which is why it wasn't used in the first place. -- Yves
    ojacobson committed with demerphq Oct 29, 2012
Commits on Dec 27, 2012
  1. Merge remote-tracking branch 'github/master'

    Merge trivial changes between git-deploy upstream and ours.
    avar committed Dec 27, 2012
Commits on Dec 23, 2012
  1. git-deploy: untabify

    avar committed Dec 23, 2012
Commits on Dec 20, 2012
  1. Tpyo

    Burak Gursoy committed Dec 20, 2012
Commits on Dec 19, 2012
  1. v6.01

        - Add a post-start hook phase where we can print out custom
          messages or end the start phase (e.g. if rollouts are blocked).
        - Pass GIT_DEPLOY_ACTION to hooks so hooks can see what action
          started something.
        - Removed specific Git::Deploy::Hook::* code.
        - Add where to get the source to the POD and be explicit about
          what license git-deploy is under.
        - Fixed an issue where we'd die when trying to print out an error
          on invalid refs.
        - Add facility to emit custom error messages on sync hook failure
          by setting deploy.message-on-sync-hook-fail-whale in the config.
        - Skip tests (instead of failing) on platforms that don't have Git
    avar committed Dec 19, 2012
  2. Changes: changes for v6.01

    avar committed Dec 19, 2012