can now unset values in CSV editing (sf bug 704788)

[roundup.git] / CHANGES.txt

This file contains the changes to the Roundup system over time. The entries
are given with the most recent entry first.

2003-??-?? 0.6.0
Removed:
- having served its purpose as a template for other relational database
  implementations, the gadfly backend has now been removed from the Roundup
  distribution.

Feature:
- support setting of properties on message and file through web and
  email interface (thanks John Rouillard)
- allow additional control over the roundupdb email sending (explicit
  cc addresses, different from address and different nosy list property)
  (thanks John Rouillard)
- applied patch for nicer history display (sf feature 638280)
- cleaning old unused sessions only once per hour, not on every cgi 
  request. It is greatly improves web interface performance, especially
  on trackers under high load
- added mysql backend (see doc/mysql.txt for details)
- switch metakit to use "compressed" multilink journal change representation
- metakit now handles "unset" for most types (not Number and Boolean)
- fixed bug in metakit search-by-ID
- added ability to display localized dates in web interface. User input is
  convered to GMT (see doc/upgrading.txt).
- added a form to show a specific issue
- more proper sorting/grouping on mulitilink properties. Sorting is performed
  not only by number of links, but also by links itself. This makes usable
  grouping e.g. by topic multilink
- add "ago" to intervals in the past (sf bug 679232)
- included UN*X manual pages from Bastian Kleineidam
- implemented extension to form parsing to allow editing of multiple items
  and creation of multiple items (but only one per class)
- the colon ":" special form variable designator may now be any of : + @
- trackers' templates directory can contain subdirectories with static files
  (e.g. images). They are accessible naturally: _file/images/img.gif
- altered Class.create() and FileClass.create() methods to make "content"
  property available in auditors
- can now configure CC to author only for messages creating issues (sf
  feature 625808)
- registration is now a two-step process, with confirmation from the email
  address supplied in the registration form
- added password reset feature for forgotten password / login
- added support for last-modified and if-modified-since headers for static
  file serving
- added Node.get() method
- nicer page titles (sf feature 65197)
- relaxed CVS importing (sf feature 693277)
- added support for searching on ranges of dates (see doc/user_guide.txt in
  chapter "Searching Page" for details)
- role names made case insensitive
- added ability to restore retired nodes


Fixed:
- applied unicode patch. All data is stored in utf-8. Incoming messages
  converted from any encoding to utf-8, outgoing messages are encoded 
  according to rfc2822 (sf bug 568873)
- fixed layout issues with forms in sidebar
- fixed timelog example so it handles new issues (sf bug 678908)
- handle missing os.fork() (sf bug 681046)
- added warning filter for "FutureWarning: hex/oct constants > sys.maxint will
  return positive values..." (literal 0xffff0000 in portalocker.py)
- fixed ZPT code generating SyntaxWarning for assignment to None
- open static files using binary mode (sf bug 693208)
- fixed deja-vu bug 692910
- don't display "Editing" on read-only pages (sf bug 651967)
- re-worked detectors initialisation - woohoo, no more cross-importing!
- fixed export/import of retired nodes (sf bug 685273)
- remember the display template specified during edit (sf bug 701815)
- added example HTML tempating for vacation flag (sf bug 701722)
- only look for CSV files when importing (thanks Dan Grassi)
- can now unset values in CSV editing (sf bug 704788)


2003-??-?? 0.5.7
- fixed Interval maths (sf bug 665357)
- fixed sqlite rollback/caching bug (sf bug 689383)
- fixed rdbms table update detection logic (sf bug 703297)


2003-02-27 0.5.6
- fixed templating filter function arguments (sf bug 678911)
- fixed multiselect in searching (sf bug 676874)
- fixed parsing of content-disposition filenames (sf bug 675116)
- added 'h' to roundup-server optarg list (sf bug 674070)
- fixed doc for db.history in anydbm and rdbms_common (sf bug 679221)
- fixed roundup-reminder (sf bug 681042)
- fixed int assumptions about Number values (sf bug 677762)
- clarified licensing
- another attempt to fix cookie misbehaviour - customise cookie name using
  tracker name
- fixed error in indexargs_url (thanks Patrick Ohly)
- fixed getnode (sf bug 684531)
- fixed args to some date templating methods (sf bug 689670)
- fixed database corruption in rdbms property mutation


2003-01-24 0.5.5
- fixed rdbms searching by ID (sf bug 666615)
- fixed metakit searching by ID
- detect corrupted index and raise semi-useful exception (sf bug 666767)
- open server logfile unbuffered
- revert StringHTMLProperty to not hyperlink text by default
- fixes to CGI form handling
- fix unlink bug in metakit backend
- fixed hyperlinking ambiguity (sf bug 669777)
- fixed cookie path to use TRACKER_WEB (sf bug 667020) (thanks Nathaniel Smith
  for helping chase it down and Luke Opperman for confirming fix)


2003-01-10 0.5.4
- key the templates cache off full path, not filename
- implemented whole-database locking
- hyperlinking of special text (url, email, item designator) in messages
- fixed time default in date.py
- fixed error in cgi/templates.py (sf bug 652089)
- fixed handling of missing password (sf bug 655632)
- applied patches for handling Outlook quirks (thanks Andrey Lebedev)
  (multipart/alternative, "fw" and content-type "name")
- fire auditors and reactors in rdbms retire (thanks Sheila King)
- better match for mailgw help "command" text
- handle :add: better in cgi form parsing (sf bug 663235)
- handle all-whitespace multilink values in forms (sf bug 663855)
- fixed searching on date / interval fields (sf bug 658157)
- fixed form elements names in search form to allow grouping and sorting 
  on "creation" field
- display of saved queries is now performed correctly


2002-12-11 0.5.3
- added mention of how to give users multiple Roles
- mention needed trailing "/" in TRACKER_WEB
- fixed upgrading doc to have CGI changes in the correct order
- fixed double-close of anydbm backend (sf bug 639030)
- removed use of string/strop from TAL/TALInterpreter
- handle KeyboardInterrupt nicely
- fixed Date and Interval form value handling
- fixed Date.local()
- email quoted text stripping is controllable again (sf bug 650742)
- extract attachment name from content-disposition if name is missing (sf
  bug 637278)
- removed FILTER_POSITION from bundled configs
- reverse message listing in issue display (reversion of recent change)
- bad entries for multilink editing in cgi don't traceback now (sf bug 640310)
- detect and break email loops (sf bug 640854)
- finished of handling of retired flag in filter() (sf bug 635260)
- allow StringHTMLProperty in MultilinkHTMLProperty test to work
- don't set explicit None Link properties in web create
- fixed nasty sorting bug that was lowercasing properties
- allow multiple :remove and :add elements per property being edited
- added date header to emails (sf bug 651358)


2002-11-07 0.5.2
- added quotes around python interpreter in windows bat (sf bug 623963)
- fixed link at end of installation doc (sf bug 623957)
- handle "classname" URL path errors cleaner (generate a 404)
- added CGI :remove:<propname> and :add:<propname> which specify item ids to
  remove / add in <propname> multilink.
- bugfix in boolean templating
- remember the change note on bad submissions (sf bug 625989)
- highlight required form fields (sf bug 625989)
- force non-word boundary to match re: in subject (sf bug 626303)
- handle sqlite bug (<2.7.2) (sf bug 630828)
- handle missing props in anydbm stringFind
- updated email package address formatting (deprecation)
- copied email address quoting from email v2.4.3 so we're consistent with 2.2
- email summary extraction now takes the first whole sentence or line -
  whichever is longer
- documented dependency on Active State (sf bug 623959)
- ensured there's no zero-length files in source (sf bug 633622)
- added ID to the search page (sf bug 631601)
- fixed filtering by id in anydbm
- show issue ID in the headings (sf bug 631598)
- show entire messages by default in issues (sf bug 625995)
- fixed journalling to save old values instead of new (sorry it took so long GM)
- handle missing REQUEST_URI for cgi-bin users (sf bug 620163)


2002-10-16 0.5.1
- highlight rows in groups of three
- metakit cleanups
- nicer "navigation" style in index views
- handle missing Link values in anydbm backend set() operation
- fixed filter() with no sort/group (sf bug 618614)
- fixed register with no session (sf bug 618611)
- fixed log / pid file path handling in roundup-server (sf bug 617981)
- fixed old gadfly compatibiltiy problem, for sure this time (sf bug 612873)
- https URLs from config now recognised as valid (sf bug 619829)
- nicer display of tracker list in roundup-server (sf bug 619769)
- fixed some missed renaming instance -> tracker (sf bug 619769)
- allow blank passwords again (sf bug 619714)
- expose the tracker config as a variable for templating
- homogenise newlines in CGI text submissions (sf bug 614072)
- merged Zope Collector #372 fix from ZPT CVS trunk
- fixed history to display username instead of userid
- shipped templates didn't import all hyperdb types in dbinit.py
- fixed bug in Interval serialisation
- handle "unset" status in status auditor (sf bug 621250)
- issues in 'done-cbb' are now also moved to 'chatting' on new messages
- implemented the missing Interval.__add__
- added ability to implement new templating utility methods
- expose the Date.pretty method to templating
- made form table cell alignment consistent (sf bug 621887)
- include stylesheet in docs (sf bug 623183)
- store PIPE messages so we can re-send them on errors (sf bug 623082)
- implemented "retire" cgi action, added to user index (sf bug 618612)
- included doc ideas from Bernhard Reiter (sf feature 621941)


2002-10-02 0.5.0
- fixed style for alternating rows in user lists
- fixed query edit form so it doesn't barf
- #617133 ] 0.5.0pr1 uses nonexistent renderTemplate
- merged Zope Collector #539 fix from ZPT CVS trunk


2002-09-27 0.5.0 pr1
- handling of None for Date/Interval/Password values in export/import
- handling of journal values in export/import
- password edit now has a confirmation field
- registration error punts back to register page
- gadfly backend now handles changes to the schema - but only one property
  at a time
- cgi.client base URL is now obtained from the config TRACKER_WEB
- request.url has gone away - there's too much magic in trying to figure
  what it should be
- cgi-bin script redirects to https now if the request was https
- FileClass "content" property wasn't being returned by getprops() in most
  backends
- we now verify instance attributes on instance open and throw a useful error
  if they're not all there
- sf 611217 ] menu() has problems when labelprop==None
- verify contents of tracker module when the tracker is opened
- many performance improvements in *dbm and sql backends
- mailgw was missing an "import sys"
- setup now installs scripts with python -O flag, doubling performance in some
  cases (there's a lot of __debug__ use)
- fix :required for Link menus
- import wasn't setting the ID to maxid+1
- added getItem to HTMLClass so you can access arbitrary items in templates
- index filtering form values may now be key values too
- replaced the content() callback ickiness with Page Template macro usage
- changed the default CSS style to be less offensive to some ;)
- better handling of Page Template compilation errors
- handle multiple unrelated indexed classes
- #614188 ] Exception in mailgw.py
- #613310 ] traceback on onexistant items
- #613291 ] typos in nosy list
- handle stupid mailers that QUOTE their Re; 'Re: "[issue1] bla blah"'
- giving a user a Role that doesn't exist doesn't break stuff any more
- revamped user guide, customisation guide, added maintenance guide
- merge Zope Collector #538 fix from ZPT CVS trunk (path expressions with a
  non-path final alternate no longer try to call a value returned by that
  alternate)
- merge Zope Collector #573 fix from ZPT CVS trunk
- merge Zope Collector #580 fix from ZPT CVS trunk
- added "crypt" password encoding and ability to set password with
  already encrypted password through roundup-admin
- fixed the mailgw so that anonymous users may still access it
- add hook to allow external password verification, overridable in the
  tracker interfaces module
- fixed login attempt by user that doesn't exist


2002-09-13 0.5.0 beta2
-  all backends now have a .close() method, and it's used everywhere
-  fixed bug in detectors __init__
-  switched the default issue item display to only show issue summary
   (added instructions to doc to make it display entire content)
-  MANIFEST.in was missing a lot of template files
-  added generic item editing
-  much nicer layout of template rendering errors
-  added context/is_edit_ok and context/is_view_ok convenience methods and
   implemented use of them in the classic template


2002-09-11 0.5.0 beta1
Fixed:
-  #576086 ] dumb copying mistake (frontends/ZRoundup.py)
-  installation instructions now mention "python2" in "testing your python".
-  made the unit tests run again - they were quite b0rken
-  #571170 ] gdbm deadlock
-  #576241 ] MultiLink problems in parsePropsFromForm
-  fixed the date module so that Date(". - 2d") works
-  web forms may now unset Link values (like assignedto)
-  cleanup: moved roundup.templatebuilder to roundup.templates.builder
-  instance __init__ no longer silently traps dbinit import errors

Feature:
-  new backend for metakit (thanks Gordon McMillan)
-  new backend for gadfly (it's as done as it's going to get)
-  further split the dbm backends from the core code, allowing easier
   non-dict-like backends (eg metakit, RDB)
-  implemented and used the new access control mechanisms (Permissions, Roles)
   (see doc/security.txt)
-  switched templating to use Zope's PageTemplates (yay!)
-  switched to sessions for web authentication
-  added Boolean and Number types
-  fixed the journal bloat
-  updated design document for new access controls
-  updated customisation document, including more examples
-  entire database export and import (incl files)
-  better mailgw help message (feature request #558562)
-  re-enabled link backrefs from messages (feature request #568714)
-  the page layout is now templatable
-  re-worked cgi interface to abstract out the explicit "issue" interface
-  have index page handle mid-page errors better so header and footer are
   still visible
-  we handle "not found", access and item page render errors better
-  fixed double-submit by having new-item-submit redirect at end
-  daemonify roundup-server (fork, logfile, pidfile)
-  modify cgitb to display PageTemplate errors better
-  rename to "instance" to "tracker"
-  have roundup.cgi pick up tracker config from the environment 
-  revamped look and feel in web interface
-  cleaned up stylesheet usage
-  several bug fixes and documentation fixes
-  added is_retired test to hyperdb.Class
-  added capability to save queries:
   - a query Class with name, klass (to search) and url (query string)
     properties
   - a Multilink to query on user called queries
   - html templates for query, and a list of queries in user.item
   - search form has Save button & name input
   - saved queries put in menu in pagehead
   - for migration, none of the above is required and old behavior preserved.
   - showquery translates search form <-> query string
-  cleaned up the indexer code:
   - it splits more words out
   - removed code we'll never use (roundup.roundup_indexer has the full
     implementation, and replaces roundup.indexer)
   - only index text/plain and rfc822/message (ideas for other text formats to
     index are welcome)
   - added simple unit test for indexer. Needs more tests for regression.
   - all String properties may now be indexed too. Currently there's a bit of
     "issue" specific code in the actual searching which needs to be
     addressed. In a nutshell:
     + pass 'indexme="yes"' as a String() property initialisation arg, eg:
           file = FileClass(db, "file", name=String(), type=String(),
               comment=String(indexme="yes"))
     + the comment will then be indexed and be searchable, with the results
       related back to the issue that the file is linked to
   - as a result of this work, the FileClass has a default MIME type that may
     be overridden in a subclass, or by the use of a "type" property as is
     done in the default templates.
   - the regeneration of the indexes (if necessary) is done once the schema is
     set up in the dbinit.
   - new "reindex" command in roundup-admin used to force regeneration of the 
     index
-  added email display function - mangles email addrs so they're not so easily
   scraped from the web
-  switched to using a session-based web login
-  made mailgw handle set and modify operations on multilinks (bug #579094)
-  fixed the journal bloat from multilink changes - we just log the add or
   remove operations, not the whole list


2002-06-24 0.4.2
Fixed:
-  Cleaned up the hyperdb unit tests.
-  Applied patch from Andrew W. Nosenko to give nicer Unauthorised message
   when anonymous user tries to edit. Should've been applied in 0.4.2pr1. Oops.
-  Added more detailed note to MIGRATION regarding the detectors changes.


2002-06-19 0.4.2pr1
Feature:
-  added a "detectors" directory for people to put their useful auditors and
   reactors in. Note - the roundupdb.IssueClass.sendmessage method has been
   split and renamed "nosymessage" specifically for things like the nosy
   reactor, and "send_message" which just sends the message.
-  link() htmltemplate function now has a "showid" option for links and
   multilinks. When true, it only displays the linked node id as the anchor
   text. The link value is displayed as a tooltip using the title anchor
   attribute.
   To use in eg. the superseder field, have something like this:
   <td>
    <display call="field('superseder', showid=1)">
    <display call="classhelp('issue', 'id,title', label='list', width=500)">
    <property name="superseder">
     <br>View: <display call="link('superseder', showid=1)">
    </property>
   </td>
-  stripping of the email message body can now be controlled through the
   config variables EMAIL_KEEP_QUOTED_TEXT and EMAIL_LEAVE_BODY_UNCHANGED.
-  all database files created are now group readable and writable.
-  added option to automatically add the authors and recipients of messages
   to the nosy lists with the options ADD_AUTHOR_TO_NOSY (default 'new') and
   ADD_RECIPIENTS_TO_NOSY (default 'new'). These settings emulate the current
   behaviour. Setting them to 'yes' will add the author/recipients to the nosy
   on messages that create issues and followup messages.
-  reverting to dates for intervals > 2 months sucks
-  changed the default message list in issues to display the message body
-  applied patch #558876 ] cgi client customization
-  split instance initialisation into two steps, allowing config changes
   before the database is initialised.
-  don't create an empty message on email issue creation if the email is empty
-  may now display additional fields in Multilink form menus
-  #541941 ] changing multilink properties by mail
-  #526730 ] search for messages capability
-  #505180 ] split MailGW.handle_Message
   - also changed cgi client since it was duplicating the functionality

Fixed:
-  stop sending blank (whitespace-only) notes
-  cleanup of serialisation for database storage
-  node ids are now generated from a lockable store - no more race conditions
-  sorting was applied to all nodes of the MultiLink class instead of
   to the nodes that are actually linked to in the "field" template
   function.  This adds about 20+ seconds in the display of an issue if
   your database has a 1000 or more issues in it.
-  added missing documentation for a few of the config option values
-  file upload broke if you didn't supply a change note
-  fixed SCRIPT_NAME in ZRoundup for instances not at top level of Zope
   (thanks dman)
-  fixed some sorting issues that were breaking some unit tests under py2.2
-  mailgw test output dir was confusing the init test (but only on 2.2 *shrug*)
-  node caching now works, and gives a small boost in performance
-  #449374 ] re-enable bsddb3 backend
   bsddb3 backend now works, reinstating
-  #551483 ] assignedto in Client.make_index_link
-  made backends.__init__ be more specific about which ImportErrors it really
   wants to ignore
-  fixed the example addresses in the templates to use correct example domains
-  cleaned out the template stylesheets, removing a bunch of junk that really
   wasn't necessary (font specs, styles never used) and added a style for 
   message content
-  build htmlbase if tests are run using CVS checkout
-  #565979 ] code error in hyperdb.Class.find
-  #565996 ] The "Attach a File to this Issue" fails
-  #564271 ] find() and new properties
-  #562130 ] cookie path generated from ZRoundup was wrong in some situations
-  remove CR characters embedded in messages (ZRoundup)
-  properly quote the email address and "real name" in all situations using the
    'email' module if it is available and 'rfc822' otherwise
-  #565992 ] if ISSUE_TRACKER_WEB doesn't have the trailing '/', add it
-  use the rfc822 module to ensure that every (oddball) email address and
   real-name is properly quoted
-  #558867 ] ZRoundup redirect /instance requests to /instance/ 
-  #569415 ] {version}
-  #569178 ] type error
   was fixed as part of the general cleanup of reactors


2002-03-25 - 0.4.1
Feature:
-  use blobfiles in back_anydbm which is used in back_bsddb.
   change test_db as dirlist does not work for subdirectories.
   ATTENTION: blobfiles now creates subdirectories for files.
-  add module blobfiles in backends with file access functions.
-  roundup db catch only IOError in getfile.
-  roundup db catches retrieving not existing files.
-  #503204 ] mailgw needs a default class
   - partially done - the setting of additional properties can wait for a
     better configuration system.
-  Alternate email addresses are now available for users. See the MIGRATION
   file for info on how to activate the feature.
-  #511168 ] Web interface: Adding new products
   Classes that don't provide template html get a default edit interface now:
   - access using the admin "class list" interface
   - limited to admin-only
   - requires the csv module from object-craft (url given if it's missing)
-  Added popup help for classes using the classhelp html template function.
   - add <display call="classhelp('priority', 'id,name,description')">
     to an item page, and it generates a link to a popup window which displays
     the id, name and description for the priority class. The description
     field won't exist in most installations, but it will be added to the
     default templates.
-  #517734 ] web header customisation is obscure
-  All messages sent to the nosy list are now encoded as
   quoted-printable before they are sent.
-  Fixed display of mutlilink properties when using the template
   functions, menu and plain.

Fixed:
-  Clean up mail handling, multipart handling.
-  respect encodings in non multipart messages.
-  makeHtmlBase: re.sub under python 2.2 did not replace '.', string.replace
   does it.
-  preamble in tepmlateBuilder mentioned htmldata
-  mailgw checks encoding on first part too.
-  #511586 ] unittest FAIL: testReldate_date
-  Added a uniquely Roundup header to email, "X-Roundup-Name"
-  All forms now have "double-submit" protection when Javascript is enabled
   on the client-side.
-  #516883 ] mail interface + ANONYMOUS_REGISTER
-  #516854 ] "My Issues" and redisplay
-  #517906 ] Attribute order in "View customisation"
-  #514854 ] History: "User" is always ticket creator
-  wasn't handling cvs parser feeding correctly
-  fixed some problems in date calculations (calendar.py doesn't handle over-
   and under-flow). Also, hour/minute/second intervals may now be more than
   99 each.
-  #527416 ] roundup-admin uses undefined value
-  #527503 ] unfriendly init blowup when parent dir
   (also handles UsageError correctly now in init)
-  #524129 ] roundup-admin gets python path wrong


2002-01-24 - 0.4.0
Feature:
-  much nicer history display (actualy real handling of property types etc)
-  journal entries for link and mutlilink properties can be switched on or 
   off
-  properties in change note are now sorted
-  you can now use the roundup-admin tool pack the database

Fixed:
-  the mail gateway now responds with an error message when invalid values 
   for arguments are specified for link or mutlilink properties
-  modified unit test to check nosy and assignedto when specified as arguments
-  handle attachments with no name (eg tnef)
-  fixed setting nosy as argument in subject line
-  fixed back_bsddb so it passed the journal tests
-  fixed status changes in mail gateway (eg. unread -> chatting)
-  we'll actually distribute the frontends directory now, as advertised...
-  handle stripping of "AW:" from subject line
-  htmltemplate list() wasn't sorting...
-  unit tests for html templating (and re-enabled the listbox field for
   multilinks)
-  allow abbreviation of "help" in admin tool too.
-  run_tests testReldate_date failed if LANG is 'german'
-  mailgw failures (unexpected ones) are forwarded to the roundup admin


2002-01-16 - 0.4.0b2
Fixed:
-  #495392 ] empty nosy -patch
-  #500574 ] messageid must have format <part1@part2>
-  fixed some problems with web editing and change detection
-  mail splitting wasn't detecting responses in the same "section" as quoted
   text
-  missed a "from i18n import _" in date.py
-  #501690 ] MIGRATION.txt incomplete
-  #502342 ] pipe interface
-  #502437 ] rogue reactor and unittest
-  re-enabled dumbdbm when using python >2.1.1 (ie 2.1.2, 2.2)
-  changed all config accesses so they access either the instance or the
   config attriubute on the db. This means that all config is obtained from
   instance_config instead of the mish-mash of classes. This will make
   switching to a ConfigParser setup easier too, I hope.
-  #502951 ] adding new properties to old database
-  #502953 ] nosy-like treatment of other multilinks
-  #503164 ] create and passwords
-  plain rendering of links in the htmltemplate now generate a hyperlink to
   the linked node's page.
-  #503330 ] ANONYMOUS_REGISTER now applies to mail
-  #503353 ] setting properties in initial email
-  #502956 ] filtering by multilink not supported
-  #503340 ] creating issue with [asignedto=p.ohly]
-  #502949 ] index view for non-issues and redisplay
-  #503793 ] changing assignedto resets nosy list
-  lots of date/interval related changes:
   - more relaxed date format for input
   - handle None for date/interval properties


2002-01-08 - 0.4.0b1
Feature:
-  Added INSTANCE_NAME to configuration - used in web and email to identify
   the instance.
-  Added EMAIL_SIGNATURE_POSITION to indicate where to place the roundup
   signature info in e-mails.
-  Some more flexibility in the mail gateway and more error handling.
-  Login now takes you to the page you back to the were denied access to.
-  Admin user now can has a user index link on their web interface.
-  We now have basic transaction support. Information is only written to
   the database when the commit() method is called. Only the anydbm and
   bsddb3 backends are modified in this way - the bsddb3 backend needs a
   lot more work anyway...
    - the CGI and mailgw automatically commit() at the end of processing a
      single transaction
    - the admin tool requires an explicit "commit" - it will prompt at exit
      if there are unsaved changes. A "rollback" removes all changes made
      during the session (up to the last commit).
-  Added the "display" command to the admin tool - displays a node's values
-  Message author's name appears in From: instead of roundup instance name
   (which still appears in the Reply-To:)
-  Added a Zope frontend for roundup.
-  Centralised the python version check code, bumped version to 2.1.1 (really
   needs to be 2.1.2, but that isn't released yet :)
-  much better attaching of erroneous messages in the mail gateway
-  #496356 ] Use threading in messages
   This adds the tracking of messages by message-id and allows threading
   using in-reply-to. Most e-mail clients support threading using this
   feature, and we hope to add support for it to the web gateway.

Fixed:
-  Lots of bugs, thanks Roché and others on the devel mailing list!
-  login_action and newuser_action return values were being ignored
-  Woohoo! Found that bloody re-login bug that was killing the mail
   gateway.
-  Fixed login/registration forwarding the user to the right page (or not,
   on a failure)
-  We now use weakrefs in the Classes to keep the database reference, so
   the close() method on the database is no longer needed.
-  #487480 ] roundup-server
-  #487476 ] INSTALL.txt
-  #489760 ] [issue] only subject
-  fixed doc/index.html to include the quoting in the mail alias.
-  fixed the backends __init__ so we can pydoc the backend modules
-  web i/f reports "note added" if there are no changes but a note is entered
-  we were assuming database files created by anydbm had the same name, but
   this is not the case for dbm. We now perform a much better check _and_
   cope with the anydbm implementation module changing too!
-  envelope-from is now set to the roundup-admin and not roundup itself so
   delivery reports aren't sent to roundup (thanks Patrick Ohly)
-  #495400 ] entering blanks
   Values with spaces are now accepted in roundup-admin - check the long help
   for details.
-  #496360 ] table width does not work
-  detectors were being registered multiple times
-  added tests for mailgw


2001-11-23 - 0.3.0 
Feature:
-  #467129 ] Lossage when username=e-mail-address
-  #473123 ] Change message generation for author
-  MailGW now moves 'resolved' to 'chatting' on receiving e-mail for an issue.
-  Added Structured Text rendering to htmltemplate, thanks Brad Clements.
-  Added CGI configuration via env vars (see roundup.cgi for details)
-  "roundup.cgi" is now installed to "<python-prefix>/share/roundup/cgi-bin"
-  roundup-admin now accepts abbreviated commands (eg. l = li = lis = list)
-  roundup-mailgw now supports unix mailbox and POP as sources of mail.
-  roundup-admin now handles all hyperdb exceptions
-  users may attach files to issues (and support in ext) through the web now
-  incorporated patch from Roch'e Compaan implementing attachments in nosy
   e-mail
-  added a target version field to the extended issue schema
-  added dummy hooks for I18N and some preliminary (test) markup of
   translatable messages

Fixed:
-  Fixed a bug in HTMLTemplate changes.
-  'unread' to 'chatting' automagic status change was b0rken.
-  Anonymous user lockout wasn't working.
-  roundup-server now works on Windows, thanks Juergen Hermann.
-  Fixed install documentation, also thanks Juergen Hermann.
-  Fixed some URL issues in roundup.cgi, again thanks Juergen Hermann.
-  bug #475347 ] WindowsError still not caught (patch from Juergen Hermann)
-  bug #474749 ] indentations lost
-  bug #477104 ] HTML tag error in roundup-server
-  bug #477107 ] HTTP header problem
-  bug #477687 ] conforming html
-  bug #474372 ] Netscape 4.77 do not render Support form
-  bug #477685 ] base64.decodestring breaks
-  bug #477837 ] lynx does not like the cookie
-  bug #477892 ] Password edit doesn't fix login cookie
-  newuser_action now presents error messages rather than tracebacks.
-  bug #479511 ] mailgw to pop
-  bug #479508 ] roundup-admin crash on wrong class
-  bad error report in hyperdb
-  roundup.mailgw now handles errors on the set() and create() at the end
   of processing
-  roundup.mailgw also handles messages that are passed to it that don't
   contain a From: line - apparently some POP servers can do this. It punts
   an error message to the roundup admin.
-  fixed nosy reaction and author copy handling
-  errors in nosy reaction will be propogated now (were effectively being
   squashed)
-  re-open the database as the author in mail handling
-  missing "return" in filter_section (thanks Roch'e Compaan)


2001-10-23 - 0.3.0 pre 3
Feature:
-  MailGW now moves 'unread' to 'chatting' on receiving e-mail for an issue.
-  feature #473127: Filenames. I modified the file.index and htmltemplate
   source so that the filename is used in the link and the creation
   information is displayed.
 Admin Tool (roundup-admin):
 -  Interactive mode for running multiple (independant at present) commands.
 -  Tabular display of nodes.
 -  Import and export via colon-separated files.

Changed:
-  re-organised the html templating code. Fixed some bugs, probably
   introduced some more.  Hopefully not too many.

Fixed:
-  Stand-alone server now has a configurable setuid user.
-  CGI interface wasn't handling checkboxes at all.
-  Fixed quopri usage in mailgw from bug reports on mailing list.
-  Remove the "freshen" command from the roundup-admin tool.
-  Catch errors in login - no username or password supplied.
-  Fixed editing of password (Password property type) thanks Roch'e Compaan.
-  Fixed grouping of non-str properties thanks Roch'e Compaan.
-  bug #473121: The customisation view and filters (CGI interface view
   customisation section may now be hidden (patch from Roch'e Compaan.)
-  bug #473122: Issue id sorting (hyperdb sorts strings-that-look-like-numbers
   as numbers now.
-  bug #473124: UI inconsistency with Link fields.
   This also prompted me to fix a fairly long-standing usability issue -
   that of being able to turn off certain filters.
-  bug #473125: Paragraph in e-mails
-  bug #473126: Sender unknown
-  bug #473130: Nosy list not set correctly


2001-10-11 - 0.3.0 pre 2
Fixed:
-  Hyperdatabase was inserting empty strings instead of None for missing
   property values. This broke a lot of things.


2001-10-10 - 0.3.0 pre 1
Feature:
-  roundup-admin create now prompts for property info if none is supplied
   on the command-line.
-  hyperdb Class getprops() method may now return only the mutable
   properties.
-  CGI interfaces now generate a top-level index of their known instances.

Changed:
-  Login now uses cookies, which makes it a whole lot more flexible. We can
   now support anonymous user access (read-only, unless there's an
   "anonymous" user, in which case write access is permitted). Login
   handling has been moved into cgi_client.Client.main()
-  The "extended" schema is now the default in roundup init.
-  The schemas have had their page headings modified to cope with the new
   login handling. Existing installations should copy the interfaces.py
   file from the roundup lib directory to their instance home.
-  Passwords are now encoded by default (except exising databases which
   will only be encoded when the passwords are changed). The scheme used
   at the moment is SHA - but the code is flexible enough to take any
   number of encoding systems.
-  The roundup-admin tool always operates as the "admin" user now. Database
   protection should be achieved using file system protections (see the
   documentation for details.)

Fixed:
-  Incorrectly had a Bizar Software copyright on the cgitb.py module from
   Ping - has been removed.
-  Pretty time interval wasn't handling > 1 month properly.
-  Generation of links to Link/Multilink in indexes. (thanks Hubert Hoegl)
-  AssignedTo wasn't in the "classic" schema's item page.
-  Fixed a whole bunch of places in the CGI interface where we should have
   been returning Not Found instead of throwing an exception.
-  Fixed a deviation from the spec: trying to modify the 'id' property of
   an item now throws an exception.
-  The plain() template function now html-escapes the content.
-  Change message was stuffing up for multilinks with no key property.



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

2001-08-30 - 0.2.8
Fixed:
-  Wasn't handling unguessable mime types for file uploads.
-  Missing import in mailgw.


2001-08-29 - 0.2.7
Feature:
-  Text searches are now case insensitive. All forms of text search use
   regular expressions now.

Fixed:
-  Had another 2.1-ism in the unit tests
-  Made the mail parser a little more robust w.r.t missing Subject:
   (both thanks Mikhail Sobolev)
-  Missed some isFooType usages (thanks Mikhail Sobolev for spotting them)
-  Reverted back to sending change messages to the web editor of a node so
   that the change note message is actually genrated.
-  CGI interface wasn't generating correct change messages.
-  Notes entered during a change are saved to the messages list even if
   there's no nosy list. No message is generated if there's no nosy list and
   there's no change note (since it would just duplicates the journal).
-  Completely removed the bsddb3 module from the tests - will be reinstated
   when the http://bsddb.sourceforge.net/'s bugs #439959 and #456408 are
   dealt with. One is fixed in CVS, the other pending.


2001-08-08 - 0.2.6
Note:
-  Roundup is now released under the same terms as the Python License.

Feature:
-  Added tests for instance initialisation. No more releasing the software
   with bugs in roundup.init!
-  Now bundling unittest with the package so that python 2.0 users can use
   the tests.
-  Much better error handling and messages generated by the mail gateway.

Fixed:
-  Implemented correct mail splitting. Added unit tests. Also snips
   signatures now too.
-  Bug #447671 - typo in roundup/init.py
-  Changed date.Date to use regular string formatting instead of strftime -
   win32 seems to have problems with %T and no hour... or something...
-  Bug #448484 - now catching correct exception from makedirs.
-  Instances are now opened by a special function that generates a unique
   module name for the instances on import time.


2001-08-03 - 0.2.5
Note:
-  The bsddb3 module has a bug that renders it non-functional. Users should
   select the anydbm or bsddb backend instead.

Fixed:
-  Python 2.0 does not contain the unittest module. The setup.py module now
   checks for unittest before attempting to run the unit tests.


2001-08-03 - 0.2.4
Features:
-  Added ability for cgi newblah forms to indicate that the new node
   should be linked somewhere.
-  Added time logging and file uploading to the templates.
-  Added "My Issues" and "My Support" to extended template. Changed "Your
   Details" to "My Details". Changed the "New Foo" links to "Add Foo".
   Added links for unassigned support and issues. Generally reorganised and
   cleanup the header up.
-  Changed the order of the information in the message generated by web edits.
-  Extended the range of intervals that are pretty-printed before actual dates
   are displayed.
-  Added more BUILD instructions including the "clean" command to force
   rebuild.
-  Web edit messages aren't sent to the person who did the edit any more. No
   message is generated if they are the only person on the nosy list.
-  Roundupdb now appends "mailing list" information to its messages which
   include the e-mail address and web interface address. Templates may
   override this in their db classes to include specific information (support
   instructions, etc).

Fixed:
-  Argument handling for the roundup-admin find command.
-  Handling of summary when no note supplied for newblah. Again.
-  Detection of no form in htmltemplate Field display.
-  Checklist html template command was setting wrong name.
-  2.1-specific gmtime() (no arg) call in roundup.date. (thanks Paul Wright)
-  mailgw was making naughty assumptions about the schema of the classes it
   was creating nodes for.
-  remove the $Foo$ from the HTML files stored in the htmlbase modules.
-  Instance import now imports the instance using imp.load_module so that
   we can have instance homes of "roundup" or other existing python package
   names.


2001-07-30 - 0.2.3
Big change:
-  I've split off the support class from the issue class in "extended".
   Anyone who has any support entries, sorry. It should be possible to
   write a scipt that moves the entries over pretty easily. If this causes
   you pain, I'll do so. You'll want to update your instance with the new
   code in "extended" either way.

Features:
-  Added the unit tests to the start of setup.py so they're run whenever
   we do anything distutils'y.
-  Added nicer prompting to the roundup-admin "init" command.
-  Actually, the roundup-admin code is totally revamped, and has command
   help and better command-line arg handling.
-  The cgi_client.Client base class now reflects the structure of "classic"
   rather than "extended" since "classic" is more of a "base" template.
-  Added more DB to test. Skips tests where imports fail.

Fixed:
-  One of the tests in test_date had the wrong expected result.
-  Fixed IssueClass so that superseders links to its classname rather than
   hard-coded to "issue".
-  templatebuilder was catching IOError instead of OSError.
-  The cgi_client newblah method wasn't detecting the __note form field
   properly.
-  The History command in htmltemplate didn't handle a new node (None
   nodeid) properly.


2001-07-29 - 0.2.2
Features:
-  Added implementation.txt to the doc directory. Contains implementation
   notes specific to this implementations of Roundup.
-  Cleaned up mailgw some (subclass Message for getPart) and added some
   tests for multipart splitting.
-  Better checking for html dir in templatebuilder.
-  Base hyperdb.Class now fakes the "id" property.
-  Made the classic roundup look more like the original prototype.
-  Made cgi_client and templating slightly more generic.
-  Moved some code around in cgi_client allowing for subclassing to change
   behaviour.
-  Added the fabricated property "id" to all hyperdb classes.
-  Cleanup of the link label generation (new method on hyperdb.Class to do
   it).

Fixed:
-  Everything uses errno module now to check errno values.
-  New issue form handles lack of note better now.
-  HTML templating uses section-bar style for index group headers now.
-  Fixed problem in link display when Link value is None.
-  Form handling in cgi client wasn't propogating through the previous
   query elements.
-  Fixed sort arguments generated for column headings so sorting can be
   changed now.


2001-07-28 - 0.2.1
Features:
-  Added docstring to roundup package so pydoc reports useful information.
-  Added the roundup 1 software carpentry submission HTML to the doc
   directory as "overview.html".

Fixes:
-  Fixed bug in init command - templatebuilder was assuming existence of
   "html" directory in instance home.
-  Fixed INSTALL.txt to reflect some changes in the installation and test
   procedure. Whatdya know, "setup.py install" does the script install.
   There you go...
-  Fixed some non-string node ids in cgi_client now that the hyperdb is
   strict about such things.

2001-07-26 - 0.2.0
Features:
-  Major reorganisation of code to allow multiple roundup instances and a
   single, site-packages -based installation. Also allows multiple database
   back-ends.
-  Moved the bin/ proggies into the top dir, so that it all works
   out-of-the-box
-  Added the "classic" template - a direct implementation of the Roundup
   spec. Well, as close as we're going to get, anyway.
-  Added an issue priority of support to "extended"
-  Added command-line arg handling to roundup-server so it's more useful 
   out-of-the-box.
-  Added distutils-style installation of "lib" files.
-  Added some unit tests.

Fixes:
-  Fixed bug in re generation in the filter
-  Fixed handling of None String property in grouped list headings
-  Fixed adding new issue with no change note
-  Fixed values in text input fields which contained quotes (") are now
   quoted.
-  Fixed a bug in the hyperdb filter - wrong variable names in the error
   message.

2001-07-19 - 0.1.3
-  Reldate now takes an argument "pretty" - when true, it pretty-prints the
   interval generated up to 5 days, then pretty-prints the date of last
   activity. The issue index and item now use the pretty format.
-  Classes list for admin user in CGI interface.
-  Made the view configuration more accessible, neater and more realistic.
-  Fixed list view grouping handling grouping by a Multilink or String or Link
   value of None or Date, ...  (mind you, sorting by Date???)
-  Fixed bug in the plain formatter when a Link was None.
-  Fixed ordering of list view column headings.
-  Fixed list view column heading sort links - and limited the number of
   columns to sort by to 2.
-  Added searching by glob to StringType filtering -
    ^text  - search for text at start of fields
    text$  - search for text at end of fields
    ^text$ - exactly match text in fields
    te*xt  - search for text matching "te"<any characters>"xt"
    te?xt  - search for text matching "te"<any one character>"xt"
-  Added more fields to the issue.filter and issue.index templates


2001-07-18 - 0.1.2
-  Set default index to ?:group=priority&:columns=activity,status,title so
   the priority column isn't displayed.
-  Thanks Anthony:
   - added notes to the README about Python prerequisites
   - added check to roundup.py, roundup.cgi, server.py and roundup-mailgw.py
     for python 2+ - and made the file itself parseable by 1.5.2 ;)
   - python 2.0 didn't have the default args for the time module functions.
   - better handling of db directory in initDB
-  Sorting on the extra properties defined by roundupdb classes was broken
   due to the caching used. May now sort on activity and creation
   properties, etc.
-  Set the default index to sort on activity


2001-07-18 - 0.1.1
-  Initial version release with consent of Roundup spec author, Ka-Ping Yee:
   "Amazing!  Nice work.  I'll watch for the source code on your website."

2001-07-11 - 0.1.0
-  Needed a bug tracking system. Looked around. Tried to install many
   Perl-based systems, to no avail. Got tired of waiting for Roundup to be
   released. Had just finished major product project, so needed something
   different for a while. Roundup here I come...