~loggerhead-team/loggerhead/trunk-rich

What's changed in loggerhead?

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

dev [future]

---------------

    - Make sure that binary files aren't annotated. (Martin Albisetti,

      #258848)

    - Loggerhead now serves bzr branches over HTTP and exposes the URL

      to branch them. Addresses bug #240577. (Jonathan Lange)

    - Leading blank lines in commit messages no longer result in an

      empty summary. (Colin Watson)

    - Added syntax highlighting to annotate view using

      python-pygments.  Partially addresses bug #306631. (Peter Bui)

    - Convert newlines in commit messages to HTML line breaks for

      annotate and changelog views.  Addresses bug #273688. (Peter

      Bui)

    - serve-branches now errors if run behind a proxy without

      paste.deploy installed. (Michael Hudson)

    - Loggerhead should now handle file and directory names that need

      URL escaping without crashing.

    - The start-loggerhead script properly sets the wsgi.url_scheme

      from the server.webpath option. (neror, #260547)

    - The revision page defaults to unified style again, and can

      convert to a side-by-side view using JavaScript. (Michael Hudson)

    - Finish converting JavaScript from MooTools to YUI 3.

    - Leading blank lines in commit messages no longer result in an

      empty summary. (Colin Watson)

    - The user-confusing "next" and "previous" links now read "older"

      and "newer" respectively. (Michael Hudson, #297930)

    - The annotate view now contains line number anchors. (Michael

      Hudson)

    - Fix inventory pages using "//" in links. (Michael Hudson, #329668)

    - Fix problems viewing files and directories containing spaces and

      other funny characters. (Peter Bui)

    - Changelog messages are now displayed with newlines preserved.

      (Peter Bui)

    - Offer a link to see the full file diffs for a file path. (Michael

      Hudson, #333797)

    - Loggerhead can be installed as a Bazaar plugin and run by

      'bzr serve --http'. (Martin Pool)

    - Fix HTML content of source files being displayed unescaped when

      Pygments was unavailable. (Michael Hudson, #344970)

    - Fix serve-branches's path argument. (Michael Hudson, #353230)

    - serve-branches now has an option, --use-cdn, to load YUI from

      Yahoo!'s CDN. (Matt Nordhoff)

    - Fix certain race conditions for loading bzr-search. (Robert

      Collins, #334250)

    - Fix errors when using serve-branches --log-folder or --user-dirs.

      (It was calling config.get_option() incorrectly.) (Matt Nordhoff,

      bug #361238)

    - Move some caching from RAM to the disk, and other caching and

      memory usage improvements. (Michael Hudson)

    - Add a --cache-dir option to serve-branches to choose where to

      place the SQL cache, and only create one temporary SQL dir per

      process. (Matt Nordhoff, #358322)

    - Replace homebrew memory profiling code with Dozer. (Paul Hummer)

    - Use the branch's public_branch as the default suggested URL to

      branch from (Matt Nordhoff, #369767)

    - Fix a file descriptor leak (Matt Nordhoff, #370845)

    - Use transport API internally, so it is possible to specify a remote 

      URL to serve-branches. (Jelmer Vernooij, #371787)

1.10 [22Dec2008]

---------------

    - Add startup deamon script for Linux (Marius Kruger)

    - Switch navigation from file_ids to paths. Fixes bugs #260363,

      #269365 and #128926. (Martin Albisetti)

    - Fix bug #258710 ("the /files page explodes in an empty branch").

      Also minor improvements to the /files and /changes pages.

      (Marius Kruger)

    - Added --port, --host and --prefix options to serve-branches

      script. (Martin Albisetti)

    - Fixed broken template for project browsing with start-loggerhead

      (Martin Albisetti)

    - Added --reload options to restart the application when a python

      file change. (Guillermo Gonzalez)

    - Added error handling middleware. (Guillermo Gonzalez)

    - Fix bug #243415 ("Tracebacks go to console but not log

      file"). Also minor improvements to logging in serve-branches and

      start-loggerhead. (Guillermo Gonzalez)

1.6 [15Aug2008]

----------------

    - Download a diffs between revisions. (Martin Albisetti)

    - Modified templates to make loggerhead's theme easier to

      change. (Paul Hummer)

    - Default sqlite interface is now sqlite3. (Robert Collins)

    - New ninja theme sponsored by Canonical (Martin Albisetti)

    - Added COPYING file and clarified copyright headers (John Arbash Meinel)

    - Remove the .py extension requiered by the Debian Policy.

      (Jelmer Vernooij)

    - New startup script serve-branches will serve Loggerhead without

      the need of configuration, and allow you to browse through directories

      and branches. (Michael Hudson)

    - Loggerhead is no longer a TurboGears application, but rather a

      WSGI application built using Paste (see http://wsgi.org/ and

      http://pythonpaste.org/ for more about WSGI and Paste).

    - URLs now use revision numbers instead of revision ids (Martin Albisetti)

    - The scripts no longer insist on Python 2.4 -- loggerhead works

      fine with 2.5.

    - Bazaar as of version 1.5 has improved in performance enough that

      the revision cache no longer gave any noticeable benefit, so it

      was removed (the files-changed cache is still useful).

    - The templates were rewritten in Zope's TAL markup, powered by

      the simpleTAL library -- improving both the performance and

      memory consumption of rendering by a factor of around 3 for

      large pages over the old Kid templates.

    - Loggerhead's poorly performing text index was disabled. bzr-search

      is now used if the plugin is installed and a text index is present

      on the branch being viewed. (Martin Albisetti, Robert Collins).

    - Loggerhead no longer depends on bzrlib functions deprecated in

      Bazaar 1.5 and removed in 1.6 (Martin Albisetti).

    - The daemonization code was made more regular, fixing bugs

      #139161 ("Starting loggerhead process may not close its stdin

      and stdout properly") and #211526 ("Codebrowse log directory has

      unnecessarily permissive permissions")

    - Some confusion about what the 'file_id' query argument means was

      cleared up: filter_file_id now means "filter revisions to those

      that affect this file" in all views and file_id means "examine

      this file/directory" in the annotate and inventory views.

    - Dates are present more compactly.

    - The loggerhead.conf file can specify which network interface to

      bind to (Mattias Eriksson)

1.2.1  [06mar2008]

------------------

    - The changelog view was out of order when not using the revision

      cache.

1.2  [04mar2008]

    - Michael Hudson <michael.hudson@canonical.com> has mostly taken

      over the maintenance of loggerhead.

    - loggerhead now has a simple test suite -- use 'nosetests' in the

      loggerhead directory to run it.

    - The rendering performance of pages that display large amounts of

      text was improved by a factor of 4 or so.

    - loggerhead no longer caches the text of the diffs between

      revisions.  It rarely helped and wasted a lot of disk space.

    - The layout of most pages was made more "tight" to waste less

      screen real estate, and some other UI improvements (Kent

      Gibson).

    - Much dead code was removed.

    - Loggerhead now computes the files changed between revisions only

      when it needs to know this.  This is a fairly expensive

      operation in Bazaar, and doing it less massively improves

      performance and memory usage in some situations.

    - Loggerhead now takes a read lock on the branch being viewed for

      the duration of each request, which improves performance

      significantly.

    - Loggerhead no longer uses the corruption-prone Berkely DB-backed

      shelve module for its caches, and rather (ab)uses a sqlite

      database instead.

    - The creation of History objects is much faster for large

      branches (Aaron Bentley).

    - Allow the construction of URLs using revnos and file paths as

      well as revids and fileids (bug #98826):

      - For changes pages, append the newest revno to display to the

        URL, like http://.../changes/<revno>

      - For annotate pages, append the revno to display to the URL,

        followed by the path, like http://.../annotate/<revno>/<path>

      - For file listing and revision pages append the revno to

        display to the URL, like http://.../files/<revno>

      Loggerhead still generates URLs using revision and file ids for

      the moment.

    - Many bugs were fixed:

      - Loggerhead does not escape special characters when embedding a

        revision ID in a URL (bug #88286)

      - Improved robustness in the face of ghosts.

      - Don't crash on displaying a commit with an empty message (bug

        #86247)

      - codebrowse fails with infinite redirections (James Henstridge,

        bug #89854)

      - Loggerhead fails to browse revisions that change binary files

        (James Henstridge, bug #91686)

      - Loggerhead atom feeds expose internal hostname (James

        Henstridge, bug #93585)

      - loggerhead don't like page break character (0x0C) in text

        files (bug #113313)

      - codebrowse source listings don't contain line number anchors

        (bug #98826)

      - only serve up unescaped user content with "Content-Disposition:

        attachment"

      - viewing the file listing consumes a lot memory (bug #116869)

      - loggerhead can't handle empty branches (bug #119228)

      - upgrading the format of a branch behind loggerhead could make

        it fail for that branch (bug #118673)

      - Error parsing non-ascii content (bug #117799)

      - Loggerhead failed on whitespace-only commit messages.

      - Links to diffs from within a revision do not work (bug

        #119422)

      - UTF-8 patches served as ISO-8859-1, fixed by served bundles as

        application/octet-stream (bug #121336)

      - TurboGears was turning query arguments into unicode, and bzr

        stopped accepting unicode revids (bug #175228)

1.1.1  [24jan2007]

------------------

    - fix broken inventory page (oops!)

    - fix a few rendering problems with kid and safari

1.1  [20jan2007]

----------------

    - new feature to compare two revisions to each other

    - inserted text in diffs is now blue instead of green

    - fixed to start and stop as a daemon now (use "-f" to run in the

      foreground), and all config is in loggerhead.conf (no need to mess with

      dev.cfg)

    - renamed show/hide javascript buttons to expand/collapse, and made them

      much faster

    - added an atom-feed link to each branch on the browse page [elliot

      murphy]

    - auto-publish feature for multiple branches under a single folder (see

      loggerhead.conf.example)

    - added the ability to share cache files per-project instead of just

      per-branch

    - added side-by-side diff display for the revision page (which is the

      default), with a button to switch between side-by-side and unified diff

      format

    - made caching use file locking, and close cleanly on shutdown

    - miscellaneous speed and page-size improvements

1.0  [23dec2006]

----------------

    - initial release