Skip to content


Subversion checkout URL

You can clone with
Download ZIP
Commits on Aug 16, 2011
  1. @chipx86

    Bump to RBTools 0.4.0 dev.

    chipx86 authored
  2. @chipx86

    Merge branch 'release-0.3.x'

    chipx86 authored
  3. @chipx86

    Bump to RBTools 0.3.4 dev.

    chipx86 authored
  4. @chipx86

    Release RBTools 0.3.3.

    chipx86 authored
  5. @chipx86

    Modernize the release script.

    chipx86 authored
    The release script is now on par with Review Board's. It generates sha1s of
    the released files and uploads them, and it knows how to update
    with the new release information.
  6. @chipx86

    Merge branch 'release-0.3.x'

    chipx86 authored
  7. @flowblok @chipx86

    Accept a single revision ID for --revision-range for Mercurial.

    flowblok authored chipx86 committed
    Mercurial now supports --revision-range with a single revision ID now. If
    given a single revision, it will find the first parent of that revision and
    use that as the first revision of the diff.
    Reviewed at
  8. @dcestari @chipx86

    Fix an error when posting a review inside a mercurial branch.

    dcestari authored chipx86 committed
    The _get_top_and_bottom_outgoing_revs method for the MercurialClient
    was assuming all revision numbers were consecutive which is not necessarily
    true at all times, especially when branches exists. This change tries to
    fetch the "parent" revision instead of just subtracting one to the bottom_rev
    to be reviewed, if this revision is not found it falls back to the old
    Reviewed at
  9. @dcestari @chipx86

    Fix an error when posting a review inside a mercurial branch.

    dcestari authored chipx86 committed
    The _get_top_and_bottom_outgoing_revs method for the MercurialClient
    was assuming all revision numbers were consecutive which is not necessarily
    true at all times, especially when branches exists. This change tries to
    fetch the "parent" revision instead of just subtracting one to the bottom_rev
    to be reviewed, if this revision is not found it falls back to the old
    Reviewed at
  10. @mbait @chipx86

    Fix Mercurial test failure related to missing config loading.

    mbait authored chipx86 committed
    A Mercurial test wasn't calling load_config_files, causing a test failure.
    Reviewed at
Commits on Aug 14, 2011
  1. @chipx86

    Fix setting the configs variable in post-review.

    chipx86 authored
    The config loading wasn't setting the global 'configs' variable. Now it's
    explicitly setting the global variable.
  2. @mbait @chipx86

    Move config loading into rbtools.utils and add tests.

    mbait authored chipx86 committed
    The code for handling config files has been moved out of post-review
    and into rbtools.utils.filesystem where it can be used by other
    Reviewed at
  3. @mbait @chipx86

    Add tests and a test base class for rbtools.utils.

    mbait authored chipx86 committed
    This introduces some unit tests for the rbtools.utils module. As part
    of this, in order to ease testing, a RBTestBase class was added, which
    contains several utility functions that will help in testing the
    client and API support.
    Reviewed at
  4. @mbait @chipx86

    Backport change: Some cleanups in the 'api' branch.

    mbait authored chipx86 committed
    Many of these cleanups apply nicely to master, and will make merging
    easier later on.
    Original description:
    Some cleanups in the 'api' branch.
    This cleans up some of the code on the 'api' branch. It removes trailing
    whitespace, aligns parameters, fixes some PEP-8 problems, gets rid of
    some old files, fixes references to new ones, and more.
    Reviewed at
Commits on Jul 26, 2011
  1. @mbait @chipx86

    Pull out code of SCM clients from postreview into separate files

    mbait authored chipx86 committed
    This is the initial work for making RBTools an extensible set of tools with
    a Python API. The various SCMClients have been pulled out into their own
    files, as have many of the utility functions that exist in post-review used
    by the clients.
    The client design is much the same as it was, with their own options and
    config data associated with the client, instead of being a global part of
    This should in the future allow for third parties to provide custom clients
    without as much work, and is an important stepping stone to merging in the
    new API work.
    Reviewed at
  2. @chipx86

    Fix breakages in the unit tests.

    chipx86 authored
    My change from a while back to add support for multiple configs ended
    up breaking unit tests, and due to other test infrastructure problems,
    was never caught. This boiled down to not properly resetting the config
    data, not loading in the config files after writing a .reviewboardrc file,
    and not defining a 'repository_url' default for OptionsStub. The unit tests
    now pass.
Commits on Jul 8, 2011
  1. @joedrago @davidt

    SVN Changelist Support

    joedrago authored davidt committed
    Add support for SVN changelists via a --svn-changelist flag.
    Based off of the change at
  2. @stodge @davidt doesn't allow parameters for http username/password

    stodge authored davidt committed can't be used by scripts that access Review Board protected
    by basic http authentication. The user is forced to manually enter a
    username and password.
    Tested on my local machine to generate review requests using an SVN hook.
    Fixes bug 1514.
    Reviewed at
Commits on Jul 3, 2011
  1. @voberoi @davidt

    Use --repository_url to override the git origin URL

    voberoi authored davidt committed
    This patch allows the user to override the current git repository's origin URL
    by setting --repository_url.
    We use this where I work with a custom ReviewBoard/Github:FI integration. We
    have canonical repos in Github that contributors fork. Contributors work in
    their fork while a script on the ReviewBoard server clones the canonical repo
    and adds a remote/fetches commits for every fork so that ReviewBoard has access
    to all commits.
    The repository URL that our integration needs in order to use post-review is the
    canonical repo, and not the fork. --repository-url seems like a reasonable entry
    point to add this functionality.
    In use in our organization by our engineering group.
    Reviewed at
  2. @davidt

    Lines starting with '---', '+++' aren't necessarily diff control lines.

    Josh Hamacher authored davidt committed
    We ran into this bug with a file containing a database schema.  It was well
    commented, so a lot of lines started with '-- ' followed with comment text.
    Unfortunately, when these lines were deleted the resulting raw diff contained
    '--- ', and as the raw diff was processed those lines would hit this code.
    We were able to submit a properly-formed diff of the schema file to ReviewBoard
    with this patch applied.  Without it, the diff was malformed.
    Reviewed at
Commits on Jun 4, 2011
  1. @dcestari @chipx86

    Add support for --guess-summary and --guess-description on Mercurial.

    dcestari authored chipx86 committed
    This adds support for the --guess-summary and --guess-description for
    Mercurial repositories when a revision range isn't specified. This wasn't
    working before for non-revision ranges.
    Reviewed at
  2. @jankoprowski @chipx86

    Support bare repositories and more revision ranges in post-review.

    jankoprowski authored chipx86 committed
    post-review couldn't be used before in bare repositories, as it expected
    a valid, full clone. Now it can be used in bare repositories, improving
    usefulness in situations where full clones aren't available and post-commit
    hooks are being used.
    Furthermore, --revision-range= can now be passed any valid git revisions
    that git rev-parse understands.
    Reviewed at
  3. @davegxmos @chipx86

    Add automatic parent diff support for git reviews with revision-range.

    davegxmos authored chipx86 committed
    When using a revision-range, you need to specify a parent diff if the first
    revision is not available in the upstream which reviewboard can see (e.g. via
    gitweb).  Hence I have made this happen automatically by detecting whether the
    upstream contains that revision and then creating a parent diff from the
    merge-base to the first revision.
    Previously it appeared that I had to do this by creating diffs manually and
    uploading them to the web interface, so this improves matters somewhat for me.
    Reviewed at
Commits on May 4, 2011
  1. @sihde @chipx86

    Teach post-review to use correct SVN remote tracking branch for git-s…

    sihde authored chipx86 committed
    …vn repositories
    When posting a diff against an SVN repository from a local git-svn repo,
    post-review always uses "master" as the parent branch.  This strategy breaks
    down in a number of cases:
    1. If your master branch has any local commits (post-review fails with a
       mysterious error message)
    2. If your master branch is behind your feature branch (e.g. you git svn
       rebase'd your feature branch more recently than your master branch)
    3. If your master branch is git svn rebase'd to a different SVN tracking branch
       than your feature branch
    4. If your local changes are on the master branch instead of a feature branch
    This change lets post-review figure out the correct SVN remote tracking branch
    (using "git svn rebase -n") and uses that as the parent; this accommodates all
    the cases listed above.
    In general it should no longer be necessary to specify --parent, though of
    course the option still works.
    Reviewed at
Commits on Apr 28, 2011
  1. @holdenk @davidt

    Make rbtools work with rb 1.5.1 with http basic auth

    holdenk authored davidt committed
    Make rbtools postreview work with 1.5.1 with HTTP basic auth.
    The changes to make this work:
        * Allow auth retries when the URL has changed
        * Allow HTTP 401 to occur when getting root /api
    Tested against internal review board install
    Reviewed at
Commits on Apr 27, 2011
  1. @kfjquinn @davidt

    RBTools Python 2.4 compatibility

    kfjquinn authored davidt committed
    Changed code syntax to avoid potential python 2.4 incompatibility.
    Tested on localhost, no errors found.
    Reviewed at
Commits on Apr 4, 2011
  1. @jankoprowski @davidt

    [RBTools] Brand new Clear Case implementation

    jankoprowski authored davidt committed
    This is new, wrote from scratch Clear Case implementation which provide new
    functionality and fix old issues:
    * recognize view type
    * remove necessity of using xargs trick to send review
    * pre-review all CHECKEDOUT files in view - just run "post-review" command
    * post-review all changes made on branch passed as --tracking-branch
    * post-review custom revision range passed as --revision-range
    * remove useless --label option
    * simple command "post-review" at least generate diff of CHECKEDOUT files
    * fixed "\No new line at end of file" workaround
    * diffs generate properly under windows
    * no cygwin required
    * Make use of diff instead broken unified_diff Python implementation where it is
    * important
    * Handle binary files
    * Updated documentation
    Tested on Linux.
    Fixes bugs 1375, 1516, 1553, 1588, 1599, 1644, 1727, 1926, 1930, 1931
    Reviewed at
Commits on Mar 31, 2011
  1. @joshk0 @davidt

    Allow post-review to work against msysgit

    joshk0 authored davidt committed
    Allow post-review to work against msysgit
    When trying to post a review from a git repository on Windows using
    msysgit, post-review will not pick up git because msysgit installs a
    'git.cmd' into the user's PATH. But post-review uses subprocess.Popen to
    launch commands, which on Windows delegates to CreateProcess, which does
    not automatically append .cmd extensions, only .exe (see MSDN for an
    explanation of this behavior.) However, CreateProcess will resolve
    things in the PATH, though, so subprocess.Popen('git.cmd', ...) does
    The fix is to try running 'git.cmd' explicitly, only on Windows.
    Testing done: Still works with msysgit and Cygwin Git.
    Fixes bug 1917
    Reviewed at
Commits on Mar 20, 2011
  1. @davidt

    Improve compatibility with older versions of Git.

    davidt authored
    Versions of Git < 1.6 use a different output format for the fetch URL of remote
    branches, and our regex wasn't catching it. This is adapted from Chris Clark's
    change at
Commits on Feb 21, 2011
  1. @chipx86

    Fix several Pygments warnings.

    chipx86 authored
    Pygments caught a bunch of unused variables and, in some cases, undefined
    variable usage. Fortunately, these were in functions that we don't actually
    call yet (http_delete), but we will in time.
Commits on Feb 20, 2011
  1. @eean @davidt

    Use real URLs when using git prefixes

    eean authored davidt committed
    So people can make arbitrary prefixes for remote URLs with Git. For example, if
    you add the following to your ~/.gitconfig :
     [url "git://"]
         insteadOf = kde:
     [url ""]
         pushInsteadOf = kde:
    Then you can git clone kde:amarok
    Currently post-review uses the prefix URL (kde:amarok) instead of the true URL
    (git:// This is a problem because people can make the
    prefix be whatever they want. This patch makes it use the true fetch URL.
    Testing done:
    Successfully resolved a prefix and also made this review (with no prefix).
    Reviewed at
Commits on Feb 16, 2011
  1. @chipx86

    Fix version detection when deciding whether to force the old API for …

    chipx86 authored
    We were checking whether Review Board 1.5.2 was used in order to determine
    whether or not to use the new API when changenums were used. We were keying
    off from 1.5.2. However, I accidentally left in code that forced the
    version number to 1.5.2, which guaranteed we'd never use the new API on any
    Review Board version when changenums were involved.
    This turned out not to be a bad thing, because we also needed the check to
    cover up to, which it now does.
Commits on Feb 9, 2011
  1. @chipx86

    Release RBTools v0.3.2.

    chipx86 authored
  2. @chipx86

    Don't use Mercurial if 'hg root' fails.

    chipx86 authored
    We were assuming the current directory was a Mercurial checkout even when
    'hg root' failed. I can't find a good reason we're doing that (perhaps
    some old compatibility issue?) but it prevents falling back on other
    SCMTools if there's any valid local Mercurial user configuration.
  3. @chipx86

    Don't use the new API on RB 1.5.2 when changenums are involved.

    chipx86 authored
    Changenums are broken in the new API, so work around that by falling back
    to the old API if using RB 1.5.2 and passing in a changenum. We'll have the
    actual API fixed in 1.5.3.
    Fixes bug #1976
Something went wrong with that request. Please try again.