10.6.0 (released 28.07.2025)

Changes:
- The VirusCheck plugin is disabled
- litecoin and ltclightning added to the list of ignored schemes

Fixes
- When a http URL was redirected to a https URL, the certificate of the https
  connection was not being verified
- TestHttp.test_html_internet was broken by Python >= 3.12.6


10.5.0 (released 03.09.2024)

Features:
- ignorewarningsforurls setting to match URLs and warnings to ignore

Fixes
- Documentation updates


10.4.0 (released 11.12.2023)

Features:
- FIFOs can be used with --config and --cookiefile

Changes:
- Minimum Python version required is 3.9
- ms-windows-store added to the list of ignored schemes
- linkchecker will exit if problems with a configuration file or cookie file
  are detected on startup

Fixes
- A cookie file that could not be parsed was causing an exception
- Documentation updates


10.3.0 (released 18.09.2023)

Features:
- TextLogger message wrapping is configurable using wraplength

Changes:
- Minimum Python version required is 3.8
- HTTP redirect causes a warning, http-redirected
- Ignored warning messages are logged as information
- Installing from git archives is re-enabled
- Support for checking NNTP and Telnet links is removed

Fixes
- -p/--password was being ignored
- FTP checker was raising a TypeError
- FTP checker was ignoring maxfilesizedownload
- Documentation updates


10.2.1 (released 05.12.2022)

Fixes
- Minimum Beautiful Soup version required restored to 4.8.1
- Documentation updates


10.2.0 (released 21.11.2022)

Features:
- ignoreerrors setting to disregard errors for URLs after checking
- AnchorCheck plugin has partial support for checking local files

Changes:
- Minimum Python version required is 3.7
- PyXDG is no longer used
- setuptools and setup.py replaced with hatchling and pyproject.toml
- The application version is derived from git tags using hatch-vcs
- Binary translation catalogs are produced using polib during distribution
  package building and are now included in sdist packages
- gemini, tg (Telegram) and whatsapp added to the list of ignored schemes
- Warning url-rate-limited renamed to http-rate-limited
- maxrequestspersecond can be less than 1
- maxrequestspersecond greater than 10 is used unchanged if the HTTP server
  returns a LinkChecker response header
- When a sitemap is discovered from a robots.txt file, the robots.txt is logged
  as the sitemap parent URL

Fixes:
- Checking directories containing Unicode filenames
- Parsing srcset attributes with multiple image candidates
- resultcachesize setting was being ignored
- sitemap output when using multiple threads
- AnchorCheck plugin is re-enabled
- Multiple man page and other documentation updates


10.1.0 (released 22.12.2021)

Features:
- resultcachesize setting to specify the maximum size of the result cache
- quiet/-q also sets the application logging level to warning
- preconnect link types are checked using DNS
- Dutch (nl_NL) application translation

Changes:
- The application version is derived from git tags using setuptools_scm
- The AnchorCheck plugin is disabled
- Binary translation catalogs are not included with the source; if present,
  polib is used by setup.py to compile translations
- The ftp_proxy environment variable is not supported. GNOME and KDE proxy
  settings are not read
- If Requests returns a character encoding of ISO-8859-1, Beautiful Soup is
  used to determine the encoding of a page; robots.txt are assumed to be UTF-8
- The linkchecker command is generated using an entry point
- GitHub Actions has replaced Travis

Fixes:
- An HTTP server can be used as an HTTPS proxy
- Multiple man page and other documentation updates


10.0.1 (released 29.1.2021)

Changes:
- Minimum supported version of Beautiful Soup is 4.8.1

Fixes:
- wsgi: Fix failure due to status logging being enabled

10.0 (released 15.1.2021)

Features:
- Uses Python 3
- C extension modules have been replaced, now uses Beautiful Soup
- Documentation converted to reStructuredText and generated with Sphinx

Changes:
- cmdline: Remove options replaced by plugins and made ineffective in 9.0
- configuration: Update proxy settings support for GNOME 3 and KDE 5
- configuration: login entries must now match the case of form element names
- logging: blacklist has been renamed to failures
- checking: Handle HTTP status code 429: Too Many Requests with
  a new warning: WARN_URL_RATE_LIMITED, instead of an error
- checking: Use timeout when fetching login forms and robots.txt
- checking: login forms with only one field are supported
- checking: slack added to the list of ignored schemes
- tests: Test coverage has been increased
- biplist is no longer used because plistlib now supports binary files
- dnspython and miniboa are no longer included
- Custom MANIFEST check replaced with check-manifest
- Code now passes flake8 checks

Fixes:
- configuration: status=0 is no longer ignored
- logging: Fix CSV logger not recognising base part setting
- logging: Fix CSV output containing increasing number of null byte characters.
- checking: Fix treating data: URIs in srcset values as links
- checking: Fix critical exception if srcset value ends with a comma
- checking: Fix critical exception when parsing a URL with a ]
- plugins: The AnchorCheck plugin is working again
- plugins: The W3C validation API has changed, CssSyntaxCheck has been updated,
  HtmlSyntaxCheck plugin is disabled
- doc: Multiple man page and other documentation updates


9.4 "just passing by" (released 12.4.2018)

Features:
- checking: Support itms-services: URLs.
  Closes: GH bug #532
- checking: Support XDG Base Directory Specification for configuration
  and data.
  Closes: GH bug #44
- add Dockerfile
- use xdg dirs for config & data
- use tox for tests and fix travis build
- add --no-robots commandline flag
- Added plugin for parsing and checking links in Markdown files

Changes:
- installation: Remove dependency on msgfmt.py by pre-generating the
  *.mo files and adding them to version control.
  Reason was the difficulty to run msgfmt.py under both Python 2 and 3.
- checking: When checking SSL certificates under POSIX systems try
  to use the system certificate store.
- logging: improved debugging by also enabling urllib3 output
- remove third-party packages and use them as dependency
- Allow wayback-format urls without affecting atom 'feed' urls
- Move dev requirements into dev-requirements.txt
- Crawl HTML attributes in deterministic order
- Remove platform-specific installer stuff and ensure a build .whl wheel file can be built.
- Move GUI files to separate project

Fixes:
- checking: Correct typos in the proxy handling code.
  Closes: GH bug #536
- checking: Add to default HTTP client headers instead of replacing.
- cmdline: Reactivate paging of help pages.
- requirements: Fix requests module version check.
  Closes: GH bug #548
- load cookies from the --cookiefile correctly
- fix incorrect call to the logging module
- Fix TypeError: hasattr(): attribute name must be string
- fix HTTPS URL checks


9.3 "Better Living Through Chemistry" (released 16.7.2014)

Features:
- checking: Parse and check links in PDF files.
- checking: Parse Refresh: and Content-Location: HTTP headers for URLs.

Changes:
- plugins: PDF and Word checks are now parser plugins
  (PdfParser, WordParser). Both plugins are not enabled
  by default since they require third party modules.
- plugins: Print a warning for enabled plugins that could not
  import needed third party modules.
- checking: Treat empty URLs as same as parent URL.
  Closes: GH bug #524
- installation: Replaced the twill dependency with local code.

Fixes:
- checking: Catch XML parse errors in sitemap XML files and print them
  as warnings. Patch by Mark-Hetherington.
  Closes: GH bug #516
- checking: Fix internal URL match pattern. Patch by Mark-Hetherington.
  Closes: GH bug #510
- checking: Recalculate extern status after HTTP redirection.
  Patch by Mark-Hetherington.
  Closes: GH bug #515
- checking: Do not strip quotes from already resolved URLs.
  Closes: GH bug #521
- cgi: Sanitize configuration.
  Closes: GH bug #519
- checking: Use user-supplied authentication and proxies when requestiong
  robot.txt.
- plugins: Fix Word file check plugin.
  Closes: GH bug #530


9.2 "Rick and Morty" (released 23.4.2014)

Fixes:
- checking: Don't scan external robots.txt sitemap URLs.
  Closes: GH bug #495
- installation: Correct case for pip install command.
  Closes: GH bug #498

Features:
- checking: Parse and check HTTP Link: headers.
- checking: Support parsing of HTML image srcset attributes.
- checking: Support parsing of HTML schema itemtype attributes.


9.1 "Don Jon" (released 30.3.2014)

Features:
- checking: Support parsing of sitemap and sitemap index XML files.
  Closes: GH bug #413
- checking: Add new HTTP header info plugin.
- logging: Support arbitrary encodings in CSV output.
  Closes: GH bug #467
- installation: Use .gz compression for source release to support
  "pip install".
  Closes: GH bug #461

Changes:
- checking: Ignored URLs are reported earlier now.
- checking: Updated the list of unkonwn or ignored URI schemes.
- checking: Internal errors do not disable check threads anymore.
- checking: Disable URL length warning for data: URLs.
- checking: Do not warn about missing addresses on mailto links that have
  subjects.
- checking: Check and display SSL certificate info even on redirects.
  Closes: GH bug #489
- installation: Check requirement for Python requests >= 2.2.0.
  Closes: GH bug #478
- logging: Display downloaded bytes.

Fixes:
- checking: Fix internal errors in debug output.
  Closes: GH bug #472
- checking: Fix URL result caching.
- checking: Fix assertion in external link checking.
- checking: Fix SSL errors on Windows.
  Closes: GH bug #471
- checking: Fix error when SNI checks are enabled.
  Closes: GH bug #488
- gui: Fix warning regex settings.
  Closes: GH bug #485


9.0 "The Wolf of Wall Street" (released 3.3.2014)

Features:
- checking: Support connection and content check plugins.
- checking: Move lots of custom checks like Antivirus and syntax
  checks into plugins (see upgrading.txt for more info).
- checking: Add options to limit the number of requests per second,
  allowed URL schemes and maximum file or download size.
  Closes: GH bug #397, #465, #420
- checking: Support checking Sitemap: URLs in robots.txt files.
- checking: Reduced memory usage when caching checked links.
  Closes: GH bug #429
- gui: UI language can be changed dynamically.
  Closes: GH bug #391

Changes:
- checking: Use the Python requests module for HTTP and HTTPS requests.
  Closes: GH bug #393, #463, #417
- logging: Removed download, domains and robots.txt statistics.
- logging: HTML output is now in HTML5.
- checking: Removed 301 warning since 301 redirects are used
  a lot without updating the old URL links.
  Also, recursive redirection is not checked any more since there
  is a maximum redirection limit anyway.
  Closes: GH bug #444, #419
- checking: Disallowed access by robots.txt is an info now, not
  a warning. Otherwise it produces a lot of warnings which
  is counter-productive.
- checking: Do not check SMTP connections for mailto: URLs anymore.
  It resulted in lots of false warnings since spam prevention
  usually disallows direct SMTP connections from unrecognized
  client IPs.
- checking: Only internal URLs are checked as default. To check
  external urls use --check-extern.
  Closes: GH bug #394, #460
- checking: Document that gconf and KDE proxy settings are parsed.
  Closes: GH bug #424
- checking: Disable twill page refreshing.
  Closes: GH bug #423
- checking: The default number of checking threads is 10 now instead of 100.

Fixes:
- logging: Status was printed every second regardless of the
  configured wait time.
- logging: Add missing column name to SQL insert command.
  Closes: GH bug #399
- checking: Several speed and memory usage improvements.
- logging: Fix --no-warnings option.
  Closes: GH bug #457
- logging: The -o none now sets the exit code.
  Closes: GH bug #451
- checking: For login pages, use twill form field counter if
  the field has neither name nor id.
  Closes: GH bug #428
- configuration: Check regular expressions for errors.
  Closes: GH bug #410


8.6 "About Time" (released 8.1.2014)

Changes:
- checking: Add "Accept" HTTP header.
  Closes: GH bug #395

Fixes:
- installer: Include missing logger classes for Windows and
  OSX installer.
  Closes: GH bug #448


8.5 "Christmas Vacation" (released 24.12.2013)

Features:
- checking: Make per-host connection limits configurable.
- checking: Avoid DoS in SSL certificate host matcher.

Changes:
- checking: Always use the W3C validator to check HTML or CSS syntax.
- checking: Remove the http-wrong-redirect warning.
- checking: Remove the url-content-duplicate warning.
- checking: Make SSL certificate verification optional and allow
  user-specified certificate files.
  Closes: GH bug #387
- cmdline: Replace argument parsing. No changes in functionality, only
  the help text will be formatted different.
- gui: Check early if help files are not found.
  Closes: GH bug #437
- gui: Remember the last "Save result as" selection.
  Closes: GH bug #380

Fixes:
- checking: Apache Coyote (the HTTP server of Tomcat) sends the wrong
  Content-Type on HEAD requests. Automatically fallback to GET in this
  case.
  Closes: GH bug #414
- checking: Do not use GET on POST forms.
  Closes: GH bug #405
- scripts: Fix argument parsing in linkchecker-nagios
  Closes: GH bug #404
- installation: Fix building on OS X systems.


8.4 "Frankenweenie" (released 25.01.2013)

Features:
- checking: Support <link rel="dns-prefetch"> URLs.
- logging: Sending SIGUSR1 signal prints the stack trace of all current
  running threads. This makes debugging deadlocks easier.
- gui: Support Drag-and-Drop of local files. If the local file is
  a LinkChecker project (.lcp) file it is loaded, else the check
  URL is set to the local file URL.

Changes:
- checking: Increase per-host connection limits to speed up checking.

Fixes:
- checking: Fix a crash when closing a Word document after scanning failed.
  Closes: GH bug #369
- checking: Catch UnicodeError from idna.encode() fixing an internal error when
  trying to connect to certain invalid hostnames.
- checking: Always close HTTP connections without body content.
  See also http://bugs.python.org/issue16298
  Closes: GH bug #376


8.3 "Mahna Mahna Killer" (released 6.1.2013)

Features:
- project: The Project moved to Github.
  Closes: GH bug #368

Changes:
- logging: Print system arguments (sys.argv) and variable values in
  internal error information.
- installation: Install the dns Python module into linkcheck_dns subdirectory to avoid
  conflicts with an upstream python-dns installation.

Fixes:
- gui: Fix storing of ignore lines in options.
  Closes: SF bug #3587386


8.2 "Belle De Jour" (released 9.11.2012)

Changes:
- checking: Print a warning when passwords are found in the configuration file
  and the file is accessible by others.
- checking: Add debug statements for unparseable content types.
  Closes: SF bug #3579714
- checking: Turn off caching. This improves memory performance drastically
  and it's a very seldom used feature - judging from user feedback over the years
  and my own experience.
- checking: Only allow checking of local files when parent URL does not exist or
  it's also a file URL.

Fixes:
- checking: Fix anchor checking of cached HTTP URLs.
  Closes: SF bug #3577743
- checking: Fix cookie path matching with empty paths.
  Closes: SF bug #3578005
- checking: Fix handling of non-ASCII exceptions (regression in 8.1).
  Closes: SF bug #3579766
- configuration: Fix configuration directory creation on Windows
  systems.
  Closes: SF bug #3584837


8.1 "Safety Not Guaranteed" (released 14.10.2012)

Features:
- checking: Allow specification of maximum checking time or maximum
  number of checked URLs.
- checking: Send a HTTP Do-Not-Track header.
- checking: Check URL length. Print error on URL longer than 2000 characters,
  warning for longer than 255 characters.
- checking: Warn about duplicate URL contents.
- logging: A new XML sitemap logger can be used that implements the protocol
  defined at http://www.sitemaps.org/protocol.php.

Changes:
- doc: Mention 7-zip and Peazip to extract the .tar.xz under Windows.
  Closes: SF bug #3564733
- logging: Print download and cache statistics in text output logger.
- logging: Print warning tag in text output logger. Makes warning filtering
  more easy.
- logging: Make the last modification time a separate field in logging
  output. See doc/upgrading.txt for compatibility changes.
- logging: All sitemap loggers log all valid URLs regardless of the
  --warnings or --complete options. This way the sitemaps can be
  logged to file without changing the output of URLs in other loggers.
- logging: Ignored warnings are now never logged, even when the URL
  has errors.
- checking: Improved robots.txt caching by using finer grained locking.
- checking: Limit number of concurrent connections to FTP and HTTP
  servers. This avoids spurious BadStatusLine errors.

Fixes:
- logging: Close logger properly on I/O errors.
  Closes: SF bug #3567476
- checking: Fix wrong method name when printing SSL certificate warnings.
- checking: Catch ValueError on invalid cookie expiration dates.
  Patch from Charles Jones.
  Closes: SF bug #3575556
- checking: Detect and handle remote filesystem errors when checking
  local file links.


8.0 "Luminaris" (released 2.9.2012)

Features:
- checking: Verify SSL certificates for HTTPS connections. Both the
  hostname and the expiration date are checked.
- checking: Always compare encoded anchor names.
  Closes: SF bug #3538365
- checking: Support WML sites.
  Closes: SF bug #3553175
- checking: Show number of parsed URLs in page content.
- cmdline: Added Nagios plugin script.

Changes:
- dependencies: Python >= 2.7.2 is now required
- gui: Display debug output text with fixed-width font.
- gui: Display the real name in the URL properties.
  Closes: SF bug #3542976
- gui: Make URL properties selectable with the mouse.
  Closes: SF bug #3561129
- checking: Ignore feed: URLs.
- checking: --ignore-url now really ignores the URLs instead
  of checking only the syntax.
- checking: Increase the default number of checker threads from 10 to
  100.

Fixes:
- gui: Fix saving of the debugmemory option.
- checking: Do not handle <object codebase="..."> attribute as parent
  URL but as normal URL to be checked.
- checking: Fix UNC path handling on Windows.
- checking: Detect more sites not supporting HEAD requests properly.
  Closes: SF bug #3535981


7.9 "The Dark Knight" (released 10.6.2012)

Fixes:
- checking: Catch any errors initializing the MIME database.
  Closes: SF bug #3528450
- checking: Fix writing temporary files.
- checking: Properly handle URLs with user/password information.
  Closes: SF bug #3529812

Changes:
- checking: Ignore URLs from local PHP files with execution
  directives of the form "<? ?>".
  Prevents false errors when checking local PHP files.
  Closes: SF bug #3532763
- checking: Allow configuration of local webroot directory to
  enable checking of local HTML files with absolute URLs.
  Closes: SF bug #3533203

Features:
- installation: Support RPM building with cx_Freeze.
- installation: Added .desktop files for POSIX systems.
- checking: Allow writing of a memory dump file to debug memory
  problems.


7.8 "Gangster Exchange" (released 12.5.2012)

Fixes:
- checking: Always use GET for Zope servers since their HEAD support
  is broken.
  Closes: SF bug #3522710
- installation: Install correct MSVC++ runtime DLL version for Windows.
- installation: Install missing Python modules for twill, cssutils and
  HTMLTidy.

Changes:
- documentation: Made the --ignore-url documentation more clear.
  Patch from Charles Jones.
  Closes: SF bug #3522351
- installation: Report missing py2app instead of generating a
  Distutils error.
  Closes: SF bug #3522265
- documentation: Fix typo in linkcheckerrc.5 manual page.
  Closes: SF bug #3522846

Features:
- installation: Add dependency declaration documentation to setup.py.
  Closes: SF bug #3524757


7.7 "Intouchables" (released 22.04.2012)

Fixes:
- checking: Detect invalid empty cookie values.
  Patch by Charles Jones.
  Closes: SF bug #3514219
- checking: Fix cache key for URL connections on redirect.
  Closes: SF bug #3514748
- gui: Fix update check when content could not be downloaded.
  Closes: SF bug #3515959
- i18n: Make locale domain name lowercase, fixing the .mo-file
  lookup on Unix systems.
- checking: Fix CSV output with German locale.
  Closes: SF bug #3516400
- checking: Write correct statistics when saving results in the GUI.
  Closes: SF bug #3515980

Changes:
- cmdline: Remove deprecated options --check-css-w3 and
  --check-html-w3.

Features:
- cgi: Added a WSGI script to replace the CGI script.


7.6 "Trkisch fr Anfnger" (released 31.03.2012)

Fixes:
- checking: Recheck extern status on HTTP redirects even if domain
  did not change. Patch by Charles Jones.
  Closes: SF bug #3495407
- checking: Fix non-ascii HTTP header handling.
  Closes: SF bug #3495621
- checking: Fix non-ascii HTTP header debugging.
  Closes: SF bug #3488675
- checking: Improved error message for connect errors to the ClamAV
  virus checking daemon.
- gui: Replace configuration filename in options dialog.
- checking: Honor the charset encoding of the Content-Type HTTP
  header when parsing HTML. Fixes characters displayed as '?'
  for non-ISO-8859-1 websites.
  Closes: SF bug #3388257
- checking: HTML parser detects and handles invalid comments of the
  form "<! bla >".
  Closes: SF bug #3509848
- checking: Store cookies on redirects. Patch by Charles Jones.
  Closes: SF bug #3513345
- checking: Fix concatenation of multiple cookie values.
  Patch by Charles Jones.
- logging: Encode comments when logging CSV comments.
  Closes: SF bug #3513415

Changes:
- checking: Add real url to cache. Improves output for cached errors.
- checking: Specify timeout for SMTP connections. Avoids spurious
  connect errors when checking email addresses.
  Closes: SF bug #3504366

Features:
- config: Allow --pause and --cookiefile to be set in configuration file.


7.5 "Kukushka" (released 13.02.2012)

Fixes:
- checking: Properly handle non-ascii HTTP header values.
  Closes: SF bug #3473359
- checking: Work around a Squid proxy bug which resulted in not
  detecting broken links.
  Closes: SF bug #3472341
- documentation: Fix typo in the manual page.
  Closes: SF bug #3485876

Changes:
- checking: Add steam:// URIs to the list of ignored URIs.
  Closes: SF bug #3471570
- checking: Deprecate the --check-html-w3 and --check-css-w3 options.
  The W3C checkers are automatically used if a local check library
  is not installed.
- distribution: The portable version of LinkChecker does not write
  the configuration file in the user directory anymore. So a user
  can use this version on a foreign system without leaving any traces
  behind.

Features:
- gui: Add Ctrl-L shortcut to highlight the URL input.
- gui: Support loading and saving of project files.
  Closes: SF bug #3467492


7.4 "Warrior" (released 07.01.2012)

Fixes:
- gui: Fix saving of check results as a file.
  Closes: SF bug #3466545, #3470389

Changes:
- checking: The archive attribute of <applet> and <object> is a
  comma-separated list of URIs. The value is now split and each URI
  is checked separately.
- cmdline: Remove deprecated options.
- configuration: The dictionary-based logging configuration is now
  used. The logging.conf file has been removed.
- dependencies: Python >= 2.7 is now required

Features:
- checking: Add HTML5 link elements and attributes.


7.3 "Attack the block" (released 25.12.2011)

Fixes:
- configuration: Properly detect home directory on OS X systems.
  Closes: SF bug #3423110
- checking: Proper error reporting for too-long unicode hostnames.
  Closes: SF bug #3438553
- checking: Do not remove whitespace inside URLs given on the
  commandline or GUI. Only remove whitespace at the start and end.
- cmdline: Return with non-zero exit value when internal program
  errors occurred.
- gui: Fix saving of check results as a file.

Changes:
- gui: Display all options in one dialog instead of tabbed panes.

Features:
- gui: Add configuration for warning strings instead of regular
  expressions. The regular expressions can still be configured in
  the configuration file.
- gui: Add configuration for ignore URL patterns.
  Closes: SF bug #3311262
- checking: Support parsing of Safari Bookmark files.


7.2 "Drive" (released 20.10.2011)

Fixes:
- checking: HTML parser now correctly detects character encoding for
  some sites.
  Closes: SF bug #3388291
- logging: Fix SQL output.
  Closes: SF bug #3415274, #3422230
- checking: Fix W3C HTML checking by using the new soap12 output.
  Closes: SF bug #3413022
- gui: Fix startup when configuration file contains errors.
  Closes: SF bug #3392021
- checking: Ignore errors trying to get FTP feature set.
  Closes: SF bug #3424719

Changes:
- configuration: Parse logger and logging part names case insensitive.
  Closes: SF bug #3380114
- gui: Add actions to find bookmark files to the edit menu.

Features:
- checking: If a warning regex is configured, multiple matches in
  the URL content are added as warnings.
  Closes: SF bug #3412317
- gui: Allow configuration of a warning regex.


7.1 "A fish called Wanda" (released 6.8.2011)

Fixes:
- checking: HTML parser detects and handles stray "<" characters before
  end tags.
- checking: Reset content type setting after loading HTTP headers again.
  Closes: SF bug #3324125
- checking: Remove query and fragment parts of file URLs. Fixes false
  errors checking sites on local file systems.
  Closes: SF bug #3308753
- checking: Do not append a stray newline character when encoding
  authentication information to base64. Fixes HTTP basic
  authentication.
  Closes: SF bug #3377193
- checking: Ignore attribute errors when printing the Qt version.
- checking: Update cookie values instead of adding duplicate entries.
  Closes: SF bug #3373910
- checking: Send cookies in as few headers as possible.
  Closes: SF bug #3346972
- checking: Send all domain-matching cookies that apply.
  Closes: SF bug #3375899
- gui: Properly reset active URL count when checking stops.
  Closes: SF bug #3311270

Changes:
- gui: Default to last URL checked in GUI (if no URL is given as
  commandline parameter).
  Closes: SF bug #3311271
- cgi: Removed FastCGI module. The normal CGI module should be
  sufficient.
- doc: Document the list of supported warnings in the linkcheckerrc(5)
  man page.
  Closes: SF bug #3340449

Features:
- checking: New option --user-agent to set the User-Agent header
  string sent to HTTP web servers. Note that this does not change
  or prevent robots.txt checking.
  Closes: SF bug #3325026


7.0 "Plots with a View" (released 28.5.2011)

Fixes:
- doc: Correct reference to RFC 2616 for cookie file format.
  Closes: SF bug #3299557
- checking: HTML parser detects and handles stray "<" characters.
  Closes: SF bug #3302895
- checking: Correct wrong import path in configuration file.
  Closes: SF bug #3305351
- checking: Only check warning patterns in parseable content.
  Avoids false errors downloading large binary files.
  Closes: SF bug #3297970
- checking: Correctly include dns.rdtypes.IN and dns.rdtypes.ANY
  submodules in Windows and OSX installers. Fixes false DNS errors.
  Closes: SF bug #3297235

Changes:
- gui: Display status info into GUI main window instead of modal window.
  Closes: SF bug #3297252
- gui: Display warnings in result column.
  Closes: SF bug #3298036
- gui: Improved option dialog layout.
  Closes: SF bug #3302498
- doc: Document the ability to search for URLs with --warning-regex.
  Closes: SF bug #3297248
- checking: Support for a system configuration file has been removed.
  There is now only one user-configurable configuration file.
- doc: Paginate linkchecker -h output when printing to console.

Features:
- logging: Colorize number of errors in text output logger.
- checking: Support both Chromium and Google Chrome profile dirs
  for finding bookmark files.
- gui: Remember last 10 checked URLs in GUI.
  Closes: SF bug #3297243
- gui: Display the number of selected rows as status message.
  Closes: SF bug #3297247


6.9 "Cowboy Bebop" (released 6.5.2011)

Fixes:
- gui: Correctly reset logger statistics.
- gui: Fixed saving of parent URL source.
- installer: Fixed portable windows version by not compressing DLLs.
- checking: Catch socket errors when resolving GeoIP country data.

Changes:
- checking: Automatically allow redirections from URLs given by the
  user.
- checking: Limit download file size to 5MB.
  SF bug #3297970
- gui: While checking, show new URLs added in the URL list view by
  scrolling down.
- gui: Display release date in about dialog.
  Closes: SF bug #3297255
- gui: Warn before closing changed editor window.
  Closes: SF bug #3297245
- doc: Improved warningregex example in default configuration file.
  Closes: SF bug #3297254

Features:
- gui: Add syntax highlighting for Qt editor in case QScintilla
  is not installed.
- gui: Highlight check results and colorize number of errors.
- gui: Reload configuration after changes have been made in the editor.
  Closes: SF bug #3297242


6.8 "Ghost in the shell" (released 26.4.2011)

Fixes:
- checking: Make module detection more robust by catching OSError.

Changes:
- gui: Print detected module information in about dialog.
- gui: Close application on Ctrl-C.
- checking: Ignore redirections if the scheme is not HTTP,
  HTTPS or FTP.
- build: Ship Microsoft C++ runtime files directly instead
  of the installer package.
- gui: Make QScintilla editor optional by falling back to a
  QPlainText editor.

Features:
- build: Support building a binary installer in 64bit Windows
  systems.
- build: The Windows installer is now signed with a local self-signed
  certificate.
- build: Added a Mac OS X binary installer.
- network: Support getting network information on Mac OS X systems.


6.7 "Friendship" (released 12.4.2011)

Fixes:
- gui: Fix display of warnings in property pane.
  Closes: SF bug #3263974
- gui: Don't forget to write statistics when saving result files.
- doc: Added configuration file locations in HTML documentation.
- doc: Removed mentioning of old -s option from man page.
- logging: Only write configured output parts in CSV logger.
- logging: Correctly encode CSV output.
  Closes: SF bug #3263848
- logging: Don't print empty country information.
- gui: Don't crash while handling internal error in non-main threads.

Changes:
- gui: Improved display of internal errors.
- logging: Print more detailed locale information on internal
  errors.

Features:
- gui: Added CSV output type for results.
- gui: Use Qt Macintosh widget style on OS X systems.
- logging: Print recursion level in machine readable logger outputs
  xml, csv and sql. Allows filtering the output by recursion level.


6.6 "Coraline" (released 25.3.2011)

Fixes:
- gui: Really read system and user configuration file.
- gui: Fix "File->Save results" command.
  Closes: SF bug #3223290

Changes:
- logging: Add warning tag attribute in XML loggers.

Features:
- gui: Added a crash handler which displays exceptions
  in a dialog window.


6.5 "The Abyss" (released 13.3.2011)

Fixes:
- checking: Fix typo calling get_temp_file() function.
  Closes: SF bug #3196917
- checking: Prevent false positives when detecting the MIME type
  of certain archive files.
- checking: Correct conversion between file URLs and encoded
  filenames. Fixes false errors when handling files with Unicode
  encodings.
- checking: Work around a Python 2.7 regression in parsing certain
  URLs with paths starting with a digit.
- cmdline: Fix filename completion if path starts with ~
- cgi: Prevent encoding errors printing to sys.stdout using an
  encoding wrapper.

Changes:
- checking: Use HTTP GET requests to work around buggy IIS servers
  sending false positive status codes for HEAD requests.
- checking: Strip leading and trailing whitespace from URLs and print
  a warning instead of having errors.
  Also all embedded whitespace is stripped from URLs given at the
  commandline or the GUI.
  Closes: SF bug #3196918

Features:
- configuration: Support reading GNOME and KDE proxy settings.


6.4 "The Sunset Limited" (released 20.2.2011)

Fixes:
- checking: Do not remove CGI parameters when joining URLs.
- checking: Correctly detect empty FTP paths as directories.
- checking: Reuse connections more than once and ensure they are
  closed before expiring.
- checking: Make sure "ignore" URL patterns are checked before
  "nofollow" URL patterns.
  Closes: SF bug #3184973
- install: Properly include all linkcheck.dns submodules in the
  .exe installer.
- gui: Remove old context menu action to view URL properties.
- gui: Disable viewing of parent URL source if it's a directory.

Changes:
- gui: Use Alt-key shortcuts for menu entries.
- checking: Improved thread locking and reduce calls to time.sleep().
- cmdline: Deprecate the --priority commandline option. Now the check
  process runs with normal priority.
- cmdline: Deprecate the --allow-root commandline option. Root
  privileges are now always dropped.
- cmdline: Deprecate the --interactive commandline option. It has
  no effect anymore.

Features:
- checking: Added support for Google Chrome bookmark files.
- gui: Preselect filename on save dialog when editing file:// URLs.
  Closes: SF bug #3176022
- gui: Add context menu entries for finding Google Chrome and Opera
  bookmark files.


6.3 "Due Date" (released 6.2.2011)

Fixes:
- install: Fixed the install instructions.
  Closes: SF bug #3153484
- logging: Enforce encoding error policy when writing to stdout.
- checking: Prevent error message from Geoip by using the correct
  API function when no city database is installed.
- checking: Properly detect case where IPv6 is not supported.
  Closes: SF bug #3167249

Changes:
- gui: Detect local or development versions in update check.


6.2 "Despicable Me" (released 6.1.2011)

Changes:
- checking: Parse PHP files recursively.
- gui: Remove reset button from option dialog.

Features:
- gui: Add update check for newer versions of LinkChecker.


6.1 "Christmas Vacation" (released 23.12.2010)

Fixes:
- checking: Fix broken anchor checking.
  Closes: SF bug #3140765
- checking: Properly detect filenames with spaces as
  internal links when given as start URL.
- logging: Allow Unicode strings to be written to stdout without
  encoding errors on Unix systems.
- logging: Fix missing content type for cached URLs.
- gui: Reset statistics before each run.

Changes:
- install: Compress Windows installer with upx, saving some Bytes.

Features:
- gui: Add URL input context menu action to paste Firefox bookmark file.
- install: Added a portable package for Windows.


6.0 "Kung Fu Panda Holiday Special" (released 19.12.2010)

Fixes:
- checking: Fall back to HTTP GET requests when the connection has
  been reset since some servers tend to do this for HEAD requests.
  Closes: SF bug #3114622
- gui: Activate links in property dialog.
- gui: Fix sorting of columns in URL result list.
  Closes: SF bug #3131401
- checking: Fix wrong __init__ call to URL proxy handler.
  Closes: SF bug #3118254
- checking: Catch socket errors (for example socket.timeout)
  when closing SMTP connections.

Changes:
- dependencies: Require and use Python 2.6.
- cmdline: Removed deprecated options --no-anchor-caching and
  --no-proxy-for.
- config: Remove backwards compatilibity parsing and require the
  new multiline configuration syntax.
- logging: Use codecs module for proper output encoding.
  Closes: SF bug #3114624
- checking: The maximum file size of FTP files is now limited
  to 10MB.
- checking: Remove warning about using Unicode domains which are more
  widely supported now.
- logging: The unique ID of an URL is not printed out anymore.
  Instead the cache URL key should be used to uniquely identify URLs.
- gui: Display URL properties in main window instead of an extra
  dialog.

Features:
- logging: More statistic information about content types and URL
  lengths is printed out.
- gui: Store column widths in registry settings.
- gui: Add ability to save results to local files with File->Save.
- gui: Assume the entered URL starts with http:// if it has no
  scheme specified and is not a valid local file.
- gui: Display check statistics in main window.
- gui: There is now a clear button in the URL input field if any text
  has been written to it.


5.5 "Red" (released 20.11.2010)

Fixes:
- checking: Do not check content of already cached URLs.
  Closes: SF bug #1720083
- checking: Do not parse URL CGI part recursively, avoiding maximum
  recursion limit errors.
  Closes: SF bug #3096115
- logging: Avoid error when logger fields "intro" or "outro" are
  configured.
- logging: Correctly quote edge labels of graph output formats and
  remove whitespace.
- checking: Make sure the check for external domain is done after all
  HTTP redirections.
- checking: Check for allowed content read before trying to
  parse anchors in HTML file.
  Closes: SF bug #3110569

Changes:
- cmdline: Don't log a warning if URL has been redirected.
  Closes: SF bug #3078820
- checking: Do not print warnings for HTTP -> HTTPS and HTTPS -> HTTP
  redirects any more.
- logging: Changed comment format in GML output to be able to load the
  graph in gephi.
- gui: Remove timeout and thread options.
- checking: Do not report irc:// hyperlinks as errors, ignore them
  instead.
  Closes: SF bug #3106302

Features:
- gui: Add command to save the parent URL source in a local file.
- gui: Show configuration files in option dialog and allow them to
  be edited.
  Closes: SF bug #3102201
- gui: Added dialog to show detailed URL properties on double click.
- gui: Store GUI options in registry settings.


5.4 "How to train your dragon" (released 26.10.2010)

Fixes:
- gui: Enable the cancel button again after it has been clicked and
  disabled.
- checking: Fix printing of active URLs on Ctrl-C.
- checking: Check for allowed content read before trying to
  parse robots.txt allowance.
- gui: Prevent off-screen window position.
  Closes: SF bug #3025284

Changes:
- gui: Display cancel message in progress window.
- gui: Use separate debug log window.
- install: Copy and execute the Microsoft Visual C runtime DLL
  installer. This solves startup error on WinXP systems that don't
  have this DLL installed.
  Closes: SF bug #3025284
- checking: Tune timeout values to close threads faster on exit.
  Closes: SF bug #3087944
- config: Authentication password entries are optional and if missing
  have to be entered at the commandline.

Features:
- gui: Added "View parent URL online" context menu action to display
  source in text editor window.
  Closes: SF bug #3040378
- gui: Read default options from configuration file.
  Closes: SF bug #2931320
- config: Added configuration file option for the --cookies command line
  option.
- http: Allow specifying a login URL in the configuration file which
  gets visited before checking submits login data.
  Closes: SF bug #3041527


5.3 "Inception" (released 29.9.2010)

Fixes:
- ftp: Fix support for FTP ports other than the default.
- build: Use _WIN32 instead of WIN32 define to detect Windows systems.
  Closes: SF bug #2978524
- http: Send correct host header when using proxies. Thanks Jason Martin
  for the patch.
  Closes: SF bug #3035754
- file: Prevent truncation of UNC paths on Windows systems.
  Closes: SF bug #3017391
- url: Work around a Python bug cutting off characters when joining an
  URL that starts with semicolon.
  Closes: SF bug #3056136
- gui: Enable tree widget items to make them selectable. This makes
  the right-click context menu work again.
  Closes: SF bug #3040377

Changes:
- checking: Caches are now size-restricted to limit the memory
  usage.
- logging: Use more memory-efficient wire-format for UrlBase,
  using __slots__.
  Closes: SF bug #2976995
- checking: Get size from Content-Length HTTP header and for local
  files from stat(2) so size information is available without
  downloading the content data.
- checking: Remove the unnormed URL warning. URLs can be written
  in more than one way and there is no norm.
  Closes: SF bug #1575800
- checking: Add "skype:" to list of ignored URL schemes.
  Closes: SF bug #2989086
- logging: Prefer the <a> element content as name instead of the title
  attribute.
  Closes: SF bug #3023483
- logging: Use semicolon as default separator for CSV files so it opens
  in Excel initially.
- checking: Allow redirections of external URLs if domain stays the
  same.
  Closes: SF bug #3024394
- cmdline: The --password option now reads a password from stdin
  instead taking it from the commandline.
- gui: Change registry base key to avoid spydoctor alert. Old keys
  have to be deleted by hand though.
  Closes: SF bug #3062161

Features:
- ftp: Detect and support UTF-8 filename encoding capability of FTP
  servers.
- checking: Added new warning to check if content size is zero.
- install: Remove Windows registry keys on uninstall.
- checking: Do not fall back to GET when no recursion is requested on
  single pages. This allows to check pages with a HEAD request even if
  robots.txt disallows to get the page content.
- checking: detect and warn when obfuscated IP addresses are found.
- gui: Add "Copy to clipboard" context menu item to copy an URL to
  the system clipboard.
- checking: Support the pygeoip package to display country information
  on windows systems.


5.2 "11:14" (released 7.3.2010)

Fixes:
- logging: Use default platform encoding instead of hardcoded one
  of iso-8859-1.
  Closes: SF bug #2770077
- dns: Use /dev/urandom instead of /dev/random to get initial seed
  on Linux machines since the last one can block indefinitely.
  Closes: SF bug #2901667
- http: Retry if server closed connection and sent an empty
  status line. Fixes the "BadStatusLine" errors.
- http: Prevent UnicodeDecodeError on redirection by ensuring that
  the redirected URL will be Unicode encoded.
- checking: Prevent UnicodeDecodeError in robots.txt parser by
  encoding the linkchecker useragent string.
- installer: Add commandline executable to Windows installer.
  Closes: SF bug #2903257
- http: Warn about permanent redirections even when redirected URL is
  outside of the domain filter.
  Closes: SF bug #2920182
- mailto: An empty email-address is syntactically allowed according
  to RFC2368. So the syntax error about missing email-addresses gets
  demoted to a warning.
  Closes: SF bug #2910588
- cmdline: Expand tilde (~) in filenames given with the --config option.

Changes:
- cmdline: disabled and deprecated the --no-proxy-for option. Use the
  $no_proxy environment variable instead.
- dns: Updated dnspython module from upstream version 1.8.1.
- checking: Improved HTML parsing speed:
  a) The parsers for HTML title and robots.txt meta tags stop after seeing
     a <body> tag.
  b) Anchor references are not always parsed, but onl when the--anchor
     option was given.
  c) Found HTML links are not queued after parsing the whole file, but
     directly when found. This also saves some memory.

Features:
- checking: Check hyperlinks of Word documents. Needs pywin32
  installed.
- http: Allow and support HTTPS proxies.


5.1 "Let the right one in" (released 04.08.2009)

Fixes:
- logging: The content size of downloads is now shown again.
- logging: The CSV logger does not crash anymore when only parts
  of log output was configured
  Closes: SF bug #2806790
- http: Fixed persistent connection handling: retry connecting to HTTP
  servers which close persistent connections unexpectedly.
- bookmarks: correctly read the bookmark title from Mozilla places.sqllite
- checking: ignore the fragment part (ie. the anchor) of URIs when
  getting and caching HTTP content; follows the HTTP/1.1 specification which
  does not include fragments in the protocol. Thanks to Martin von Gagern
  for pointing this out.
  This also deprecates the --no-anchor-caching option which will be
  removed in future releases.
  Closes: SF bug #2784996
- checking: Prefer to encode spaces with %20 instead of + to be sure mailto:
  URLs are understood by email clients.
  Closes: SF bug #2820773
- checking: Allow digits at end of domain names.

Changes:
- logging: Switch default output encoding of loggers to UTF-8, except the
  text logger which also honors the system settings.
  Closes: SF bug #2579899
- logging: Make output more concise by not logging duplicate cached URLs.
- nntp: Only retry 3 instead of 5 times to connect to busy NNTP servers.
- cmdline: The command line script exits with error only when errors
  or warnings are printed. Previously it exited with error status even
  when all warnings were ignored.
  Closes: SF bug #2820812

Features:
- email: Added email syntax checking.
  Closes: SF bug #2595437
- gui: Improved progress dialog in GUI client: show active and queued URLs.
- gui: Added right-click context menu for logged URLs.
- nntp: Output welcome message from NNTP servers as info.
- http: Honor the no_proxy environment variable.
- config: the system configuration is copied to the user configuration at
  ~./linkchecker/linkcheckerrc if it does not exist yet.
- logging: the loggers now have an additional field "ID" which prints
  a unique ID for each logged URL.

5.0.2 "All the boys love Mandy Lane" (released 13.2.2009)

  * Properly detect location of the log configuration file in the Windows
    binary .exe.
    Closes: SF bug #2564674

  * Install locale .mo files in the Windows binary .exe

5.0.1 "Slumdog Millionaire" (released 31.1.2009)

  * Remove unit tests from distribution to avoid antivirus software
    alarms with the virus filter tests.
    Closes: SF bug #2537822

  * Updated dnspython module from upstream.
    Changed: linkcheck/dns/*, tests/dns/*

5.0 "Iron Man" (released 24.1.2009)

  * Require and use Python >= 2.5.
    Type: feature
    Changed: *.py

  * Send HTTP Referer header for both http and https URLs.
    Type: feature
    Changed: linkcheck/checker/httpurl.py

  * The HTML and CSS syntax check now only applies to URLs
    which match those given on the command line.
    This makes checking of personal pages easier.
    Type: feature
    Changed: linkcheck/checker/urlbase.py

  * Added online HTML and CSS syntax checks using W3C validators.
    Implemented as commandline options --check-html-w3 and
    --check-css-w3.
    Type: feature
    Changed: linkchecker, linkcheck/checker/urlbase.py

  * Added ability to scan URL content with ClamAV virus scanner.
    Implemented as commandline option --scan-virus.
    Type: feature:
    Changed: linkchecker, linkcheck/checker/urlbase.py
    Added: linkcheck/clamav.py

  * Improved network interface detection on POSIX systems.
    Type: bugfix
    Added: linkcheck/network/*

  * Improved graph output: print labels as node names. Thanks
    to Jan Weiss for the initial idea.
    Type: feature
    Changed: linkcheck/logger/{dot,gml,gxml}.py
    Added: linkcheck/logger/graph.py

  * Add support for setuptools and thus Python eggs in the setup.py
    script. This should fix installation errors for generated .egg
    files.
    Type: feature
    Closes: SF bug #1985509

  * Support parsing of HTML pages served with application/xhtml+xml
    content type.
    Type: bugfix
    Closes: SF bug #1994104

  * Support reading URLs from stdin in the commandline interface.
    Type: feature
    Closes: SF bug #2013873, #2013874
    Changed: linkchecker

  * Improved filename recognition on Windows systems.
    Type: bugfix
    Changed: linkcheck/checker/fileurl.py

  * Fix error encoding non-ASCII robots.txt content. Makes some sites
    like wikipedia.org accessible with LinkChecker.
    Type: bugfix
    Changed: linkcheck/robotparser2.py

  * Fix off-by-one error in cookie domain matching code. Prevented
    some cookie files to work properly.
    Type: bugfix
    Changed: linkcheck/cookies.py
    Closes: SF bug #2016451

  * Improved double Ctrl-C abort on Unix and Windows platforms.
    Type: feature
    Changed: linkcheck/director/__init__.py

  * Support reading Firefox 3 bookmark files in SQLite format.
    Type: feature
    Changed: linkcheck/checker/fileurl.py

  * Handle non-Latin1 filenames when checking local directories.
    Type: bugfix
    Closes: SF bug #2093225
    Changed: linkcheck/checker/fileurl.py

  * Use configured proxy when requesting robots.txt, especially
    honor the noproxy values.
    Type: bugfix
    Closes: SF bug #2091297
    Changed: linkcheck/robotparser2.py, linkcheck/cache/robots_txt.py,
      linkcheck/checker/httpurl.py

  * Added new --complete option; making --verbose less chatty.
    Type: feature
    Closes: SF #2338973
    Changed: linkchecker, linkcheck/configuration/__init__.py

  * Remove gopher: URL checking.
    Type: feature
    Changed: linkcheck/checker/unkonwnurl.py
    Removed: linkcheck/checker/gopherurl.py

4.9 "Michael Clayton" (released 25.4.2008)

  * Parse Shockwave Flash (SWF) for URLs to check
    Type: feature
    Changed: linkcheck/checker/urlbase.py

  * Don't parse <script for=""> attributes since they specify IDs, not
    URLs.
    Type: bugfix
    Changed: linkcheck/linkparse.py

  * Fix bash filename completion script:
      - add missing COMPREPLY variable
      - support whitespace in files using "-o filenames" bash completion
        option
      - support subdirs by adding a FileCompleter argument matcher to
        optcomplete.autocomplete()
    Type: bugfix
    Changed: config/linkchecker-completion

  * Prevent unicode errors when an email address contains non-ascii
    characters.
    Type: bugfix
    Changed: linkcheck/checker/mailtourl.py

  * Workaround for buggy servers that break protocol synchronization of
    persistent HTTP connections.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py
    Closes: SF bug #1913992

  * Properly fall back to DNS A requests when no MX host could be found
    for a mailto: URL.
    Type: bugfix
    Changed: linkcheck/checker/mailtourl.py
    Closes: SF bug #1942463

  * Double Ctrl-C aborts checking immediately, without cleanup.
    Type: feature
    Changed: linkcheck/director/__init__.py
    Closes: SF bug #1720104

  * Intern patterns now accept URLs with and without "www." prefixes
    as default. This allows sites to check that use both variants.
    Type: feature
    Changed: linkcheck/checker/internpaturl.py

  * Added --check-html and --check-css options to enable HTML and CSS
    syntax checking. Uses third-party modules "tidy" and "cssutils"
    for the actual check.
    Type: feature
    Changed: linkchecker, linkcheck/checker/urlbase.py

4.8 "Hallam Foe" (released 16.12.2007)

  * Fix message typo for not disclosing information.
    Type: documentation
    Closes: SF bug #1758531
    Changed: linkcheck/director/console.py, po/de.po, po/linkchecker.pot

  * Always read the request body data on persistent HTTP connections, else
    subsequent calls will get data from the previous request.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py

  * Zope server workaround: assume missing HEAD support when receiving
    text/plain on a HEAD request. Switch to GET request in this case.
    Type: bugfix
    Closes: SF bug #1770131
    Changed: linkcheck/checker/httpurl.py

  * Prevent double encoding in HTML info output.
    Type: bugfix
    Changed: linkcheck/logger/html.py

  * Honor urllib.proxy_bypass() when ignoring proxy settings.
    This only affected Windows systems, since on other platforms
    the proxy_bypass() function always return False (on Python <= 2.5
    that is).
    Type: bugfix
    Changed: linkcheck/checker/proxysupport.py

  * Document the --configfile option in the man page.
    Type: documentation
    Changed: doc/{en,de}/linkchecker.1

  * Remove comments from CSS content before searching for links.
    Type: bugfix
    Changed: linkcheck/linkparse.py, linkcheck/checker/urlbase.py
    Closes: SF bug #1831900

  * Try to detect unkonwn URL schemes from the command line, eg. URLs
    like "rtsp://foo".
    Type: feature
    Changed: linkchecker, linkcheck/lc_cgi.py,
      linkcheck/checker/{__init__,urlbase,httpurl,unknownurl}.py

  * Fix typo in warnings and use constants for the warning strings
    to avoid this in the future.
    Type: bugfix
    Closes: SF bug #1838803
    Changed: linkcheck/checker/__init__.py

  * Make sure LinkChecker does not check paths that are not prefixed
    with the start URL.
    Type: bugfix
    Closes: SF bug #1841305
    Changed: linkcheck/checker/internpaturl.py
    Added: linkcheck/checker/test/test_internpat.py

  * Try to solve the "Too many open files" errors that users have
    encountered.
    + Ensure that the connection of a checked URL are closed after checking
      (except for reused connections in the connection pool).
    + Regularly close expired connections from the connection pool, and
      finally close all of them when the program is finished.
    Closes: SF #1758338, SF #1678055, SF #1631042
    Type: bugfix
    Changed: linkcheck/cache/connection.py, linkcheck/director/aggregator.py,
      linkcheck/checker/{httpurl,mailtourl,urlbase,ftpurl}.py
    Added: linkcheck/directory/cleanup.py

  * Add man page linkcheckerrc(5) for the configuration file format.
    Type: documentation
    Added: doc/{en,de}/linkcheckerrc.5
    Changed: doc/po4a.conf

  * Drop french translations, they are less than 20% complete for
    years now.
    Type: documentation
    Removed: doc/fr/*

  * Correct misnamed colums in create.sql script: r/*string/\1/g
    Type: bugfix
    Changed: config/create.sql
    Closes: SF #1849733

  * Improved cookie parsing:
    + Allow spaces in attribute values. Example:
      "Set-Cookie: expires=Wed, 12-Dec-2001 19:27:57 GMT"
      is now parsed correctly
    + Add an optional leading dot for domain names, and account for that
      in the domain checking routine.
    Type: feature
    Changed: linkcheck/cookies.py

  * Don't print cached errors or warnings unless verbose output is
    requested.
    Type: feature
    Changed: linkcheck/director/logger.py,
      linkcheck/logger/{__init__,html,text}.py

4.7 "300" (released 17.6.2007)

  * Mention in the documentation that --anchors enables logging of
    the anchor warning.
    Type: documentation
    Changed: linkchecker, linkcheck/doc/*/linkchecker.1

  * Make sure --anchors and --no-warnings play along in the configuration.
    Type: bugfix
    Changed: linkchecker, linkcheck/configuration/__init__.py

  * Check that charset is not None before lowering it in set_encoding().
    Type: bugfix
    Changed: linkcheck/HtmlParser/__init__.py

  * Use standard "utf-8" charset name instead of "utf8" for the XML output
    encoding. Thanks to Dan Reitano for the note.
    Type: bugfix
    Changed: linkcheck/logger/xmllog.py

  * Added "created" attribute in XML output root element.
    Added "result" attribute in XML output valid element.
    Type: feature
    Changed: linkcheck/logger/customxml.py

  * Fix printing of unicode names. Thanks to Frank Bennet for the hint.
    Type: bugfix
    Changed: linkcheck/logger/{html,text}.py

  * Deprecate gopher: URLs. They do not really exist anymore and the
    gopherlib module in Python 2.5 is deprecated and will vanish soon.
    Type: feature
    Changed: doc/*/{documentation.txt,linkchecker.1}, linkchecker

4.6 "Cars" (released 16.12.2006)

  * Fixed default config file syntax by not indenting comment lines
    Type: bugfix
    Changed: config/linkcheckerrc

  * Don't set the URL result on redirections when getting the content.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py

  * Ignore errors when opening the log file output, and display a warning
    instead.
    Type: bugfix
    Closes: SF bug #1600172
    Changed: linkcheck/logger/__init__.py
    Added: linkcheck/dummy.py

  * Added some more examples.
    Added: doc/examples/*.sh

  * Pull in changes from Python subversion repository to locally stored
    gzip and httplib modules.
    Type: bugfix
    Changed: linkcheck/{httplib2,gzip2}.py

4.5 "The Good Humor Man" (released 25.9.2006)

  * Don't ignore robots.txt entries consisting only of Allow: directives.
    Type: bugfix
    Changed: linkcheck/robotparser2.py

  * Don't rely on HTTP HEAD requests to generate the same response status
    as HTTP GET. So we have to follow redirections when using HTTP GET to
    get page contents.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py

  * Document proxy URL syntax.
    Closes: SF bug #1562129
    Type: documentation
    Changed: linkchecker, {doc,po}/{en,de}.po, doc/{en,de}/linkchecker.1

  * Print active URLs on Ctrl-C interrupt.
    Closes: SF patch #1562177
    Type: feature
    Changed: linkcheck/director/__init__.py

  * Replace all old "entry1, entry2" configuration entries with
    multiline "entry" config entry. The old syntax is still supported,
    but deprecated.
    Closes: SF patch #1562195
    Type: feature
    Changed: linkcheck/configuration/confparse.py

  * If LinkChecker was not able to spawn the initial checker and status
    threads, print an informative error instead of an internal error.
    Type: feature
    Changed: linkcheck/director/__init__.py

4.4 "Garden State" (released 16.9.2006)

  * The JavaScript URL syntax check allows now digits and underscores.
    Patch from Olivier Berger.
    Type: bugfix
    Changed: cgi-bin/lconline/check.js

  * Add "internlinks" documentation and example to the default config
    file linkcheckerrc.
    Type: documentation
    Changed: config/linkcheckerrc

  * Detect more cases when a HTTP connection cannot be reused and
    must be closed. And close response objects after usage.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py

  * Only wait before a new connection to a host, not when reusing
    a previous connection.
    Type: bugfix
    Changed: linkcheck/cache/connection.py

  * Add more infos to various HTTP errors. Don't close connection when
    the response object is still open.
    Type: feature
    Changed: linkcheck/httplib2.py

  * Ignore keyboard interrupts during shutdown.
    Type: bugfix
    Changed: linkcheck/director/__init__.py

  * Removed old Psyco references from man page and documentation.
    Type: documentation
    Changed: doc/*/linkchecker.1, doc/en/install.txt

4.3 "Brick" (released 17.8.2006)

  * Use RawConfigParser for config parsing, getting rid of the unused
    interpolation feature of the default ConfigParser.
    Type: feature
    Changed: linkcheck/configuration/confparse.py
    
  * Removed the deprecated --disable-psyco option.
    Type: feature
    Changed: linkchecker

  * Allow infinite recursion in CGI script, and add a warning about
    performance requirements.
    Type: feature
    Changed: cgi-bin/lconline/lc_cgi.html.*, linkcheck/lc_cgi.py,
      doc/en/install.txt

4.2 "V for Vendetta" (released 26.7.2006)

  * Drop privileges when running as root under Unix systems. Add
    new option --allow-root to prevent this.
    Type: feature
    Changed: linkchecker, doc/{en,de}/linkchecker.1
  
  * Don't generate empty output files, open them only when they are
    written to.
    Type: bugfix
    Changed: linkcheck/logger/{__init__,text}.py

  * Only accept ASCII in robots.txt content
    Type: bugfix
    Changed: linkcheck/robotsparser2.py

  * Fix the --profile option run.
    Type: bugfix
    Changed: linkchecker

  * Remove the psyco optimizer, it prevented Ctrl-C breaking to work
    properly.
    Type: bugfix
    Changed: linkchecker

  * Norm the base reference URL.
    Type: bugfix
    Changed: linkcheck/checker/urlbase.py

  * If default encoding cannot be determined, fall back to ASCII.
    Type: bugfix
    Changed: linkcheck/i18n.py
    Closes: SF bug #1524800

4.1 "Tsotsi" (released 29.5.2006)

  * Wait for spawned threads to finish before shutdown. Gets rid
    of exceptions during shutdown.
    Type: bugfix
    Changed: linkcheck/director/*.py

  * Every once in a while look through the URL queue and put cached
    URLs to the top. This way cached URLs will get checked more quickly.
    Type: feature
    Changed: linkcheck/cache/urlqueue.py

4.0 "Down in the Valley" (released 19.5.2006)

  * Put a name to the DOT graph output. Thanks to Peter Chiocchetti
    for noticing this.
    Type: bugfix
    Changed: linkcheck/logger/dot.py

  * Parse <!> empty SGML comments in HTML data. And build the HTML
    parser with equivalence class compression which makes it a lot
    smaller and only a little tad slower.
    Also, literal </script> is not allowed anymore in single-line
    JavaScript comments in HTML data.
    Type: feature
    Changed: linkcheck/HtmlParser/htmllex.[lc],
      linkcheck/tests/test_parser.py

  * Revamp the threading algorithm by using a URL queue, with a
    constant number of consumer threads called 'workers'.
    This fixes the remaining "dequeue mutated during iteration" errors.
    Type: feature
    Changed: *.py

  * The default intern pattern matches both http: and https: schemes
    now.
    Type: feature
    Changed: linckheck/checker/internpaturl.py

  * If the robots.txt connection times out, don't bother to check
    the URL but report an error immediately. Avoids having the
    timeout twice.
    Type: feature
    Changed: linkcheck/robotparser2.py

  * DNS lookups for HTTP links are now cached.
    Type: feature
    Changed: linkcheck/httplib2.py
    Added: linkcheck/cache/addrinfo.py

  * Added timeout value option to the configuration file.
    Type: feature
    Changed: linkcheck/configuration/confparse.py, config/linkcheckerrc

  * New option --cookiefile to set initial cookie values sent to
    HTTP servers.
    Type: feature
    Changed: linkchecker, linkcheck/configuration/__init__.py,
      linkcheck/checker/httpurl.py, linkcheck/cookies.py

  * The --pause option delays requests to the same host, and is not
    required to disable threading to do that.
    Type: bugfix
    Changed: linkcheck/cache/connection.py, linkcheck/checker/urlbase.py,
      linkcheck/directory/__init__.py

  * Honor the "Crawl-delay" directive in robots.txt files.
    Type: feature
    Changed: linkcheck/robotparser2.py, linkcheck/checker/httpurl.py,
      linkcheck/cache/robots_txt.py, linkcheck/cache/connection.py,

  * Merge IgnoredUrl and ErrorUrl into UnknownUrl. Enables caching
    on invalid URLs, plus the ability to first check for external
    URL patterns.
    Type: bugfix
    Changed: linkcheck/checker/__init__.py
    Removed: linkcheck/checker/{ignored,error}url.py
    Added: linkcheck/checker/unknownurl.py

  * Convert the "label too long" domain name parse error into
    a more friendly error message.
    Type: bugfix
    Changed: linkcheck/checker/{__init__,urlbase,httpurl,fileurl}.py,
      linkchecker

3.4 "The Chumscrubbers" (released 4.2.2006)

  * Ignore decoding errors when retrieving the robots.txt URL.
    Type: bugfix
    Changed: linkcheck/robotparser2.py

  * On HTTP redirects, cache all the encountered URLs, not just the
    initial one.
    Type: feature
    Changed: linkcheck/checker/{urlbase,httpurl,cache}.py

  * Fixed the Cookie parsing and sending.
    Type: bugfix
    Changed: linkcheck/checker/cache.py
    Added: linkcheck/cookies.py

  * The psyco optimizer now has a maximum memory limit.
    Type: feature
    Changed: linkchecker

  * The checker did not recurse into command line URLs that had upper
    case characters.
    Type: bugfix
    Changed: linkcheck/checker/__init__.py
    Closes: SF bug #1413162

  * Fix a possible thread race condition by checking the return
    value of the lock.acquire() method.
    Type: bugfix
    Changed: linkcheck/decorators.py

3.3 "Four Brothers" (released 14.10.2005)

  * Fix parsing of ignore and nofollow in configuration files.
    Type: bugfix
    Changed: linkcheck/configuration.py
    Closes: SF bug #1311964, #1270783

  * Ignore refresh meta content without a recognizable URL.
    Type: bugfix
    Changed: linkcheck/linkparse.py
    Closes: SF bug #1294456

  * Catch CGI syntax errors in mailto: URLs, and add an appropriate
    warning about the error.
    Type: bugfix
    Changed: linkcheck/checker/mailtourl.py
    Closes: SF bug #1290563

  * Initialize the i18n on module load time, so one does not have
    to call init_i18n() manually anymore. Fixes parts in the code
    (ie. the CGI script) that forgot to do this.
    Type: feature
    Changed: linkcheck/__init__.py
    Closes: SF bug #1277577

  * Compress libraries in the .exe installer with UPX compressor.
    Type: feature
    Changed: setup.py

  * Ensure that base_url is Unicode for local files.
    Type: bugfix
    Changed: linkcheck/checker/fileurl.py
    Closes: Debian bug #332870

  * The default encoding for program and logger output will be the
    preferred encoding now. It is determined from your current locale
    system settings.
    Type: feature
    Changed: linkchecker, linkcheck/checker/__init__.py,
      linkcheck/i18n.py, linkcheck/logger/__init__.py

  * Improved documentation about recursion and proxy support.
    Type: documentation
    Changed: linkchecker, doc/en/documentation.txt,
      doc/{en,de}/linkchecker.1

  * Make sure that given proxy values are reasonably well-formed.
    Else abort checking of the current URL.
    Type: feature
    Changed: linkcheck/checker/proxysupport.py

  * Correctly catch internal errors in the check URL loop, and
    disable raising certain exceptions while the abort routine finishes
    up.
    Fixes the "dequeue mutated during iteration" errors.
    Type: bugfix
    Changed: linkcheck/checker/{__init__,consumer}.py
    Closes: SF bug #1325570, #1312865, #1307775, #1292919, #1264865

3.2 "Kiss kiss bang bang" (released 3.8.2005)

  * Fixed typo in redirection handling code.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py
  
  * Handle all redirections to different URL types, not just HTTP ->
    non-HTTP.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py

  * Workaround a urllib2.py bug raising ValueError on some failed
    HTTP authorisations.
    Type: bugfix
    Closes: SF bug #1250555
    Changed: linkcheck/robotparser2.py

  * Fix invalid import in DNS resolver.
    Type: bugfix
    Changed: linkcheck/dns/resolver.py

3.1 "Suspicious" (released 18.7.2005)

  * Updated documentation for the HTML parser.
    Type: feature
    Changed: linkcheck/HtmlParser/*
    
  * Added new DNS debug level and use it for DNS routines.
    Type: feature
    Changed: linkcheck/__init__.py, doc/en/linkchecker.1,
      linkcheck/dns/{ifconfig,resolver}.py

  * Use tags for different LinkChecker warnings and allow them to
    be filtered with a configuration file entry.
    Type: feature
    Changed: linkchecker, linkcheck/checker/*.py,
      linkcheck/configuration.py

  * Add compatibility fix for HTTP/0.9 servers, from Python CVS.
    Type: bugfix
    Changed: linkcheck/httplib2.py

  * Add buffer flush fix for gzip files, from Python CVS.
    Type: bugfix
    Changed: linkcheck/gzip2.py

  * Do not cache URLs where a timeout or unusual error occurred.
    This way they get re-checked.
    Type: feature
    Changed: linkcheck/checker/{__init__, urlbase}.py

  * For HTTP return codes, try to use the official W3C name when it
    is defined.
    Type: feature
    Changed: linkcheck/checker/httpurl.py

  * Fix detection code of supported GCC command line options. this
    fixes a build error on some Unix systems (eg. FreeBSD).
    Type: bugfix
    Closes: SF bug #1238906
    Changed: setup.py

  * Renamed the old "xml" output logger to "gxml" and added a new
    "xml" output logger which writes a custom XML format.
    Type: feature
    Changed: linkchecker, linkcheck/logger/*xml*.py

  * Use correct number of checked URLs in status output.
    Type: bugfix
    Closes: SF bug #1239943
    Changed: linkcheck/checker/consumer.py

3.0 "The Jacket" (released 8.7.2005)

  * Catch all check errors, not just the ones inside of URL checking.
    Type: bugfix
    Changed: linkcheck/checker/__init__.py

  * Ensure that the name of a newly created thread is ASCII. Else there
    can be encoding errors.
    Type: bugfix
    Changed: linkcheck/strformat.py, linkcheck/checker/consumer.py,
      linkcheck/threader.py

  * Use our own gzip module to cope with incomplete gzip streams.
    Type: bugfix
    Closes: SF bug #1158475
    Changed: linkcheck/checker/httpurl.py
    Added: linkcheck/gzip2.py

  * Fix hard coded python.exe path in the batch file linkchecker.bat.
    Type: bugfix
    Closes: SF bug #1206858
    Changed: setup.py, install-linkchecker.py

  * Allow empty relative URLs. Note that a completely missing URL is
    still an error (ie. <a href=""> is valid, <a href> is an error).
    Type: bugfix
    Closes: SF bug #1217397
    Changed: linkcheck/linkparse.py, linkcheck/logger/*.py,
      linkcheck/checker/urlbase.py

  * Added checks for more <meta> URL entries, especially favicon
    check was added.
    Type: feature
    Changed: linkcheck/linkparse.py

  * Limit memory consumption of psyco optimizer.
    Type: feature
    Changed: linkchecker

  * Always norm the URL before sending a request.
    Type: bugfix
    Changed: linkcheck/checker/urlbase.py

  * Send complete email address on SMTP VRFY command. Avoids a spurious
    warning about incomplete email addresses.
    Type: bugfix
    Changed: linkcheck/checker/mailtourl.py

  * The old intern/extern URL configuration has been replaced with
    a new and hopefully simpler one. Please see the documentation on
    how to upgrade to the new option syntax.
    Type: feature
    Changed: linkchecker, linkcheck/*.py

  * Honor XHTML in tag browser.
    Type: bugfix
    Closes: SF bug #1217356
    Changed: linkcheck/linkparse.py

  * Catch curses.setupterm() errors.
    Type: bugfix
    Closes: SF bug #1216092
    Changed: linkcheck/ansicolor.py

  * Only call _optcomplete bash completion function when it exists.
    Type: bugfix
    Closes: Debian bug #309076
    Changed: config/linkchecker-completion

  * If a default config file (either /etc/linkchecker/linkcheckerrc or
    ~/.linkchecker/linkcheckerrc) does not exist it is not added to
    the config file list.
    Type: bugfix
    Changed: linkcheck/configuration.py

  * The default output encoding is now that of your locale, and not
    the hardcoded iso-8859-15 anymore.
    Type: feature
    Closes: Debian bug #307810
    Changed: linkcheck/logger/__init__.py

  * Do not generate an empty user config dir ~/.linkchecker by default,
    only when needed.
    Type: feature
    Closes: Debian bug #307876
    Changed: linkchecker

  * Redundant dot path at beginning of relative urls are now removed.
    Type: feature
    Changed: linkcheck/url.py, linkcheck/tests/test_url.py

  * Displaying warnings is now the default. One can disable warnings
    with the --no-warnings option. The old --warnings option is
    deprecated.
    Type: feature
    Changed: linkchecker, linkcheck/configuration.py

  * CGI parameters in URLs are now properly splitted and normed.
    Type: bugfix
    Changed: linkcheck/url.py

  * The number of encountered warnings is printed on program end.
    Type: feature
    Changed: linkcheck/logger/{text,html}.py

  * The deprecated --status option has been removed.
    Type: feature
    Changed: linkchecker

  * New option --disable-psyco to disable psyco compilation regardless
    if it is installed.
    Type: feature
    Changed: linkchecker

  * Since URL aliases from redirections do not represent the real
    URL with regards to warnings, the aliases are no longer cached.
    Type: bugfix
    Changed: linkcheck/checker/cache.py, linkcheck/checker/httpurl.py

  * The ignored url type honors now intern/extern filters.
    Type: bugfix
    Changed: linkcheck/checker/ignoreurl.py
    Closes: SF #1223956

2.9 "Sweat" (released 22.4.2005)

  * Use collections.deque object for incoming URL list. This is faster
    than a plain Python list object.
    Type: optimization
    Changed: linkcheck/checker/cache.py

  * Updated spanish translation, thanks to Servilio Afre Puentes.
    Type: feature
    Changed: po/es.po

2.8 "Robots" (released 8.4.2005)

  * Correct AttributeError in blacklist logger.
    Type: bugfix
    Closes: SF bug #1173823
    Changed: linkcheck/logger/blacklist.py

  * Do not enforce an optional slash in empty URI paths. This resulted
    in spurious warnings.
    Closes: SF bug #1173841
    Changed: linkcheck/url.py, linkcheck/tests/test_url.py

  * On NT-derivative Windows systems, the command line scripts is now named
    "linkchecker.bat" to facilitate execution.
    Type: feature
    Changed: setup.py, install-linkchecker.py, doc/en/index.txt

  * Use pydoc.pager() in strformat.paginate() instead of rolling out
    our own paging algorithm.
    Type: feature
    Changed: linkcheck/strformat.py

2.7 "Million Dollar Baby" (released 30.3.2005)

  * When a host has no MX record, fall back to A records as the mail
    host.
    Type: bugfix
    Changed: linkcheck/checker/mailtourl.py

  * Do not split CGI params on semicolons. This is wrong of course,
    but not supported by all servers. A later version of the CGI parser
    engine will split and re-join semicolons.
    Type: bugfix
    Changed: linkcheck/url.py

  * Make sure that URLs are always Unicode strings and not None.
    Type: bugfix
    Closes: SF bug #1168720
    Changed: linkcheck/linkparse.py, linkcheck/containers.py

  * Fix the detection of persistent HTTP connections.
    Type: bugfix
    Changed: linkcheck/checker/httpheaders.py

  * HTTP connections with pending data will not be cached.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py

  * Add all URL aliases to the URL cache to avoid recursion. This
    also changes some invariants about what URLs are expected to be
    in the cache.
    Type: bugfix
    Changed: linkcheck/checker/cache.py

2.6 "Lord of the Rings" (released 15.3.2005)

  * Run with low priority. New option --priority to run with normal
    priority.
    Type: feature
    Changed: linkchecker, linkcheck/threader.py

  * If GeoIP Python wrapper is installed, log the country name as info.
    Type: feature
    Changed: linkcheck/checker/consumer.py
    Added: linkcheck/checker/geoip.py

  * New option --no-proxy-for that lets linkchecker contact the given
    hosts directly instead of going through a proxy.
    Also configurable in linkcheckerrc
    Type: feature
    Changed: linkchecker, linkcheck/checker/proxysupport.py,
      linkcheck/configuration.py

  * Give a useful error message for syntax errors in regular expressions.
    Type: bugfix
    Changed: linkchecker, linkcheck/configuration.py

  * Accept quoted urls in CSS attributes.
    Type: bugfix
    Changed: linkcheck/linkparse.py

  * Eliminate duplicate link reporting in the link parser.
    Type: bugfix
    Changed: linkcheck/linkparse.py

  * Do not send multiple Accept-Encoding headers.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py

  * Avoid deadlocks between the cache and the queue lock.
    Type: bugfix
    Changed: linkcheck/checker/consumer.py, linkcheck/checker/cache.py
    Added: linkcheck/lock.py

  * Always reinitialize stored HTTP headers on redirects; prevents
    a false alarm about recursive redirects.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py

2.5 "Spanglish" (released 4.3.2005)

  * Added spanish translation, thanks to Servilio Afre Puentes.
    Type: feature
    Changed: po/Makefile
    Added: po/es.po

  * Ignore a missing locale/ dir and fall back to the default locale
    instead of crashing.
    Type: bugfix
    Changed: linkcheck/i18n.py

  * Since profile.py and pstats.py have been removed from some
    Python standard installations (eg. Debian GNU/Linux), make their
    usage optional.
    Using --profile without an available profile.py prints a warning
    and runs linkchecker without profiling.
    Using --viewprof without an available pstats.py prints an error
    and exits.
    Type: bugfix
    Changed: linkchecker

  * Ensure stored result, info and warning strings are always Unicode.
    Else there might be encoding errors.
    Type: bugfix
    Closes: SF bug #1143553
    Changed: linkcheck/checker/{urlbase,httpurl,ftpurl}.py,
      linkcheck/strformat.py

  * Fix -h help option on Windows systems
    Type: bugfix
    Closes: SF bug #1149987
    Changed: linkchecker

2.4 "Kitchen stories" (released 9.2.2005)

  * Work around a Python 2.4 bug when HTTP 302 redirections are
    encountered in urllib2.
    Type: bugfix
    Changed: linkcheck/robotparser2.py

  * Be sure to use Unicode HTML parser messages.
    Type: bugfix
    Changed: linkcheck/linkparse.py

  * Make sure that FTP connections are opened when they are reused.
    Else open a new connection.
    Type: bugfix
    Changed: linkcheck/checker/ftpurl.py

  * Added '!' to the list of unquoted URL path characters.
    Type: bugfix
    Changed: linkcheck/url.py, linkcheck/tests/test_url.py

  * Fix Windows path name for network paths.
    Type: bugfix
    Closes: SF bug #1117839
    Changed: linkcheck/checker/fileurl.py

  * Regularly remove expired connections from the connection pool.
    Type: feature
    Changed: linkcheck/checker/pool.py

  * Documentation and pylint cleanups.
    Type: feature
    Changed: linkcheck/*.py

2.3 "Napoleon Dynamite" (released 3.2.2005)

  * Use and require Python >= 2.4.
    Type: feature
    Changed: doc/install.txt, linkcheck/__init__.py, some scripts

  * Add square brackets ([]) to the list of allowed URL characters
    that do not need to be quoted.
    Type: bugfix
    Changed: linkcheck/url.py

  * Document the return value of the linkchecker command line script
    in the help text and man pages.
    Type: documentation
    Changed: linkchecker, doc/{en,de,fr}/linkchecker.1

  * Always write the GML graph beginning, not just when "intro" field
    is defined.
    Type: bugfix
    Changed: linkcheck/logger/gml.py

  * Added DOT graph format output logger.
    Type: feature
    Added: linkcheck/logger/dot.py
    Changed: linkcheck/logger/__init__.py, linkcheck/configuration.py,
      linkchecker

  * Added ftpparse module to parse FTP LIST output lines.
    Type: feature
    Added linkcheck/ftpparse/*
    Changed: setup.py, linkcheck/checker/ftpurl.py

  * Ignore all errors when closing SMTP connections.
    Type: bugfix
    Changed: linkcheck/checker/mailtourl.py

  * Do not list FTP directory contents when they are not needed.
    Type: bugfix
    Changed: linkcheck/checker/ftpurl.py

  * Added connection pooling, used for HTTP and FTP connections.
    Type: feature
    Added: linkcheck/checker/pool.py
    Changed: linkcheck/checker/{cache, httpurl, ftpurl}.py

  * The new per-user configuration file is now stored in
    ~/.linkchecker/linkcheckerrc.
    Type: feature
    Changed: linkchecker, linkcheck/configuration.py, doc/{de,en,fr}/*.1

  * The new blacklist output file is now stored in
    ~/.linkchecker/blacklist.
    Type: feature
    Changed: linkchecker, linkcheck/configuration.py, doc/{de,en,fr}/*.1

  * Start the log output before appending new urls to the consumer since
    this can trigger logger.new_url().
    Type: bugfix
    Changed: linkcheck/checker/{__init__, consumer}.py

  * Fix crash when using -t option.
    Type: bugfix
    Changed: linkchecker

  * Updated french translation of linkchecker, thanks to Yann Verley.
    Type: feature
    Changed: po/fr.po, doc/fr/linkchecker.1

2.2 "Cube" (released 25.01.2005)

  * CSV log format changes:
    - default separator is now a comma, not a semicolon
    - the quotechar can be configured and defaults to a double quote
    - write CSV column headers as the first data row
      (thanks to Hartmut Goebel)
    Type: feature
    Changed: linkcheck/logger/csvlog.py

  * Support bzip-compressed man pages in RPM install script.
    From Hartmut Goebel.
    Type: feature
    Changed: install-rpm.sh

  * HTML parser updates:
    - supply and use Py_CLEAR macro
    - only call set_encoding function if tag name is 'meta'
    Type: feature
    Changed: linkcheck/HtmlParser/*

  * Changed documentation format for epydoc.
    Type: documentation
    Changed: *.py

  * Fix FTP error message display crash.
    Type: bugfix
    Changed: linkcheck/checker/ftpurl.py

  * Ask before overwriting old profile data with --profile.
    Type: feature
    Changed: linkchecker

  * When searching for link names, limit the amount of data to look at
    to 256 characters. Do not look at the complete content anymore.
    This speeds up parsing of big HTML files significantly.
    Type: optimization
    Changed: linkcheck/linkparse.py

  * Support Psyco >= 1.4. If you installed older versions of Psyco,
    a warning is printed.
    Type: feature
    Changed: linkchecker, doc/install.txt

  * The build script setup.py uses -std=gnu99 when using GNU gcc compilers.
    This gets rid of several compile warnings.
    Type: feature
    Changed: setup.py

  * Correct the sent User-Agent header when getting robots.txt files.
    Added a simple robots.txt example file.
    Type: bugfix
    Changed: linkcheck/robotparser2.py
    Added: doc/robots.txt

  * Updated the included linkcheck/httplib2.py from the newest httplib.py
    found in Python CVS.
    Type: feature
    Changed: linkcheck/httplib2.py

  * Do not install unit tests. Only include them in the source distribution.
    Type: feature
    Changed: MANIFEST.in, setup.py

2.1 "Shogun Assassin" (released 11.1.2005)

  * Added XHTML support to the HTML parser.
    Type: feature
    Changed: linkcheck/HtmlParser/*

  * Support plural forms in gettext translations.
    Type: feature
    Changed: po/*.po*

  * Remove intern optcomplete installation, and make it optional to
    install, since it is only needed on Unix installations using
    bash-completion.
    Type: feature
    Changed: linkchecker, config/linkchecker-completion
    Removed: linkcheck/optcomplete.py

  * Minor enhancements in url parsing.
    Type: feature
    Changed: linkcheck/url.py

  * Sort according to preference when checking MX hosts so that
    preferred MX hosts get checked first.
    Type: bugfix
    Changed: linkcheck/checker/mailtourl.py

  * If mail VRFY command fails, print a warning message.
    Type: feature
    Changed: linkcheck/checker/mailtourl.py

2.0 "I Kina spiser de hunde" (released 7.12.2004)

  * Regenerate the HTML parser with new Bison version 1.875d.
    Also use the now supported Bison memory macros YYMALLOC and
    YYFREE.
    Type: feature
    Changed: linkcheck/HtmlParser/htmlparse.y

  * Updated installation and usage documentation.
    Type: documentation
    Changed: doc/install.txt, doc/index.txt

  * Added comment() method to loggers for printing comments.
    Type: feature
    Changed: linkcheck/logger/*.py

  * Updated and translated manpages. French translation from
    Yann Verley. German translation from me ;)
    Type: documentation
    Added: doc/de/linkchecker.de.1, doc/fr/linkchecker.fr.1
    Changed: doc/en/linkchecker.1

  * Fix mailto: URL norming by splitting the query type correctly.
    Type: bugfix
    Changed: linkcheck/url.py

  * Encode all output strings for display.
    Type: bugfix
    Changed: linkchecker

  * Accept -o option logger type as case independent string.
    Type: feature
    Changed: linkchecker

  * Internal Unicode handling fixed.
    Type: bugfix
    Changed: linkcheck/url.py, linkcheck/checker/*.py

  * Use correct FTP directory list parsing.
    Type: bugfix
    Changed: linkcheck/checker/ftpurl.py

2.0rc2 "El dia de la bestia" (released 20.11.2004)

  * encode version string for --version output
    Type: bugfix
    Closes: SF bug #1067915
    Changed: linkchecker

  * Added shell config note with --home install option.
    Type: documentation
    Closes: SF bug #1067919
    Changed: doc/install.txt

  * Recheck robots.txt allowance and intern/extern filters for
    redirected URLs.
    Type: bugfix
    Closes: SF bug #1067914
    Changed: linkcheck/checker/httpurl.py

  * Updated the warning and info messages to be always complete
    sentences.
    Type: feature
    Changed: linkcheck/checker/*.py, po/*, linkcheck/ftests/*.py,
      linkcheck/ftests/data/*.result

  * Added missing script_dir to the windows installer script.
    Use python.exe instead of pythonw.exe and --interactive option to
    call linkcheck script.
    Add Documentation link to the programs group.
    Type: bugfix
    Changed: install-linkchecker.py

2.0rc1 "The Incredibles" (released 16.11.2004)

  * Only instantiate SSL connections if SSL is supported
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py

  * Close all opened log files.
    Type: bugfix
    Changed: linkcheck/logger/*.py

  * All loggers have now an output encoding. Valid encodings are listed
    in http://docs.python.org/lib/node127.html. The default encoding is
    "iso-8859-15".
    Type: feature
    Changed: linkcheck/logger/*.py

  * The --output and --file-output parameters can specify the encoding
    now. The documentation has been updated with this change.
    Type: feature
    Changed: linkchecker, linkchecker.1

  * The encoding can also be specified in the linkcheckerrc config file.
    Type: feature
    Changed: config/linkcheckerrc

  * All leading directories of a given output log file are created
    automatically now. Errors creating these directories or opening
    the log file for writing abort the checking and print a usage mesage.
    Type: feature
    Changed: linkchecker, linkcheck/logger/__init__.py

  * Coerce url names to unicode
    Type: feature
    Changed: linkcheck/checker/__init__.py

  * Accept unicode filenames for resolver config
    Type: feature
    Changed: linkcheck/dns/resolver.py

  * LinkChecker accepts now Unicode domain names and converts them
    according to RFC 3490 (http://www.faqs.org/rfcs/rfc3490.html).
    Type: feature
    Changed: linkcheck/dns/resolver.py, linkcheck/url.py

  * Exceptions in the log systems are no more caught.
    Type: feature
    Changed: linkcheck/ansicolor.py

  * Remember a <base href=""> tag in the link parser. Saves one HTML
    parse.
    Type: feature
    Changed: linkcheck/checker/urlbase.py, linkcheck/linkparse.py

  * Optimize link name parsing of img alt tags.
    Type: feature
    Changed: linkcheck/linkname.py

  * Remove all references to the old 'colored' output logger.
    Type: documentation
    Closes: SF bug #1062011
    Changed: linkchecker.1

  * Synchronized the linkchecker documentation and the man page.
    Type: documentation
    Closes: SF bug #1062034
    Changed: linkchecker, linkchecker.1

  * Make --quiet an alias for -o none.
    Type: bugfix
    Closes: SF bug #1063144
    Changed: linkchecker, linkcheck/configuration.py,
      linkcheck/checker/consumer.py

  * Re-norm a changed file:// base url, avoiding a spurious warning.
    Type: bugfix
    Changed: linkcheck/checker/fileurl.py

  * Wrong case of file links on Windows platforms now issue a
    warning.
    Type: feature
    Closes: SF bug #1062007
    Changed: linkcheck/checker/fileurl.py

  * Updated the french translation. Thanks to Yann Verley.
    Type: feature
    Changed: po/fr.po

1.13.5 "Die Musterknaben" (released 22.9.2004)
  * Use xgettext with Python support for .pot file creation, adjusted
    developer documentation.
    Type: feature
    Changed: doc/install.txt, po/Makefile, MANIFEST.in
    Removed: po/pygettext.py, po/msgfmt.py

  * Use plural gettext form for log messages.
    Type: feature
    Changed: linkcheck/logger/{text,html}.py

  * Check if FTP file really exists instead of only the parent dir.
    Type: bugfix
    Changed: linkcheck/checker/ftpurl.py

  * Document the different logger output types.
    Type: documentation
    Changed: linkchecker, linkchecker.1

  * Recursion into FTP directories and parseable files has been
    implemented.
    Type: feature
    Changed: linkcheck/checker/ftpurl.py

1.13.4 "Shaun of the dead" (released 17.9.2004)
  * Catch HTTP cookie errors and add a warning.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py

  * fix up response page object in robots.txt parser for the upcoming
    Python 2.4 release
    Type: bugfix
    Changed: linkcheck/robotparser2.py

  * remove cached urls from progress queue, fixing endless wait for
    checking to finish
    Type: bugfix
    Changed: linkcheck/checker/consumer.py

  * updated and synchronized documentation of the man page (linkchecker.1)
    and the linkchecker --help output.
    Type: documentation
    Changed: linkchecker, linkchecker.1

1.13.3 "Fight Club" (released 10.9.2004)
  * Prevent collapsing of relative parent dir paths. This fixes false
    positives on URLs of the form "../../foo".
    Closes: SF bug #1025459
    Changed: linkcheck/url.py, linkcheck/tests/test_url.py

1.13.2 "Zatoichi" (released 8.9.2004)
  * Fix permissions of data files on install to be world readable.
    Type: bugfix
    Closes: SF bug #1022132
    Changed: setup.py

  * Fixed the SQL logger when encountering empty URLs.
    Type: bugfix
    Closes: SF bug #1022156
    Changed: linkcheck/logger/sql.py

  * Added notes about access rules for CGI scripts
    Type: documentation
    Changed: doc/install.txt

  * Updated french translation. Thanks, Yann Verley!
    Type: feature
    Changed: po/fr.po

  * initialize i18n at program start
    Type: bugfix
    Changed: linkchecker, linkcheck/lc_cgi.py

  * Make initialization function for i18n, and allow LOCPATH to override
    the locale directory.
    Type: feature
    Changed: linkcheck/__init__.py

  * Removed debug print statement when issueing linkchecker --help.
    Type: bugfix
    Changed: linkchecker

  * Reset to default ANSI color scheme, we don't know what background
    color the terminal has.
    Type: bugfix
    Closes: SF bug #1022158
    Changed: linkcheck/configuration.py

  * Reinit the logger object when config files change values.
    Type: bugfix
    Changed: linkcheck/configuration.py

  * Only import ifconfig routines on POSIX systems.
    Type: bugfix
    Closes: SF bug #1024607
    Changed: linkcheck/dns/resolver.py

1.13.1 "Old men in new cars" (released 3.9.2004)
  * Fixed RPM generation by adding the generated config file to the
    installed files list.
    Type: bugfix
    Changed: setup.py

  * Mention to remove old versions when upgrading in the documentation.
    Type: documentation
    Changed: doc/upgrading.txt, doc/install.txt

  * Fix typo in redirection cache handling.
    Type: bugfix
    Changed: linkcheck/checker/cache.py

  * The -F file output must honor verbose/quiet configuration.
    Type: bugfix
    Changed: linkcheck/checker/consumer.py

  * Generate all translation files under windows systems.
    Type: bugfix
    Changed: po/Makefile

  * Added windows binary installer script and configuration.
    Type: feature
    Changed: setup.py, setup.cfg, doc/install.txt
    Added: install-linkchecker.py

  * Do not raise an error when user and/or password of ftp URLs is not
    specified.
    Type: bugfix
    Changed: linkcheck/checker/ftpurl.py

  * honor anchor part of cache url key, handle the recursion check
    with an extra cache key
    Type: bugfix
    Changed: linkcheck/checker/{urlbase,cache,fileurl}.py

  * Support URL lists in text files with one URL per line. Empty lines
    or comment lines starting with '#' are ignored.
    Type: feature
    Changed: linkcheck/checker/fileurl.py

  * Added new option --extern-strict to specify strict extern url
    patterns.
    Type: feature
    Changed: linkchecker

  * Strip quotes from parsed CSS urls.
    Type: bugfix
    Changed: linkcheck/checker/urlbase.py

1.13.0 "The Butterfly Effect" (released 1.9.2004)
  * lots of internal code restructuring
    Type: code cleanup
    Changed: a lot

  * If checking revealed errors (or warnings with --warnings),
    the command line client exits with a non-zero exit status.
    Type: feature
    Closes: SF bug 1013191
    Changed: linkchecker, linkcheck/checker/consumer.py

  * Specify the HTML doctype and charset in HTML output.
    Type: feature
    Closes: SF bug 1014283
    Changed: linkcheck/logger/html.py

  * Fix endless loop on broken urls with non-empty anchor.
    Type: bugfix
    Changed: linkcheck/checker/httpurl.py

  * For news: or nntp: urls, entries in ~/.netrc are now ignored.
    You should give instead username/password info in the configuration
    file or on the command line.
    Type: bugfix
    Changed: linkcheck/checker/nntpurl.py

  * The HTML output shows now HTML and CSS validation links for
    the parent URL of invalid links.
    Type: feature
    Changed: linkcheck/logger/html.py

  * The status is now printed as default, it can be supressed with
    the new --no-status option.
    Type: feature
    Changed: linkchecker

  * The default recursion level is now infinite.
    Type: feature
    Changed: linkchecker

  * The 'outside of domain filter' is no more a warning but an informational
    message. A warning is inappropriate since the user is in full control
    over what links are extern or intern.
    Type: feature
    Closes: SF bug 1013206
    Changed: linkcheck/urlbase.py

  * Renamed the --strict option to --extern-strict-all.
    Type: feature
    Changed: linkchecker

  * a new cache and queueing algorithm makes sure that no URL is
    checked twice.
    Type: feature
    Changed: linkcheck/checker/cache.py

  * the given user/password authententication is now also used to
    get robots.txt files.
    Type: feature
    Changed: linkcheck/robotparser2.py, linkcheck/checker/cache.py

1.12.3 "The Princess Bride" (released 27.5.2004)
  * fall back to GET on bad status line of a HEAD request
    Type: bugfix
    Changed: linkcheck/HttpUrlData.py

  * really fall back to GET with Zope servers; fixes infinite loop
    Type: bugfix
    Changed: linkcheck/HttpUrlData.py

  * better error msg on BadStatusLine error
    Type: feature
    Changed: linkcheck/UrlData.py

  * updated optcomplete to newest upstream
    Type: feature
    Changed: linkcheck/optcomplete.py

  * also quote query parts of urls
    Type: bugfix
    Changed: linkcheck/{HttpUrlData, url}.py

  * - preserve the order in which HTML attributes have been parsed
    - cope with trailing space in HTML comments
    Type: feature
    Changed: linkcheck/parser/{__init__.py,htmllex.l}
    Added: linkcheck/containers.py

  * rework anchor fallback
    Type: bugfix
    Changed: linkcheck/HttpUrlData.py

  * move contentAllowsRobot check to end of recursion check to avoid
    unnecessary GET request
    Type: bugfix
    Changed: linkcheck/UrlData.py

1.12.2 (release 4.4.2004)
  * use XmlUtils instead of xmlify for XML quoting
    Type: code cleanup
    Added: linkcheck/XmlUtils.py
    Changed: linkcheck/StringUtil.py, linkcheck/log/XMLLogger.py

  * don't require a value anymore with the --version option
    Type: bugfix
    Changed: linkchecker

  * before putting url data objects in the queue, check if they have
    correct syntax and are not already cached
    Type: optimization
    Changed: linkcheck/{UrlData,Config}.py

  * every once in a while, remove all already cached urls from the
    incoming queue. This action is reported when --status is given.
    Type: optimization
    Changed: linkcheck/Config.py

  * both changes above result in significant performance improvements
    when checking large websites, since a majority of the links tend
    to be navigation links to already-cached pages.
    Type: note

  * updated examples and put them before options in the man page for
    easier reading
    Type: documentation
    Changed: linkchecker, linkchecker.1

  * added contact url and email to the HTTP User-Agent string, which
    gets us more accepted by some bot-blocking software; also see
    http://www.livejournal.com/bots/
    Type: feature
    Changed: linkcheck/Config.py

  * only check robots.txt for http connections
    Type: bugfix
    Changed: linkcheck/{Http,}UrlData.py
    Closes: SF bug 928895

  * updated regression tests
    Type: feature
    Changed: test/test_*.py, Makefile
    Added: test/run.sh

  * preserve the order in which HTML attributes have been parsed
    Type: feature
    Changed: linkcheck/parser/{__init__.py,htmllex.l}

  * handle and correct missing start quotes in HTML attributes
    Type: feature
    Changed: linkcheck/parser/htmllex.l

  * full parsing of .css files
    Type: feature
    Changed: linkcheck/{Http,}UrlData.py, linkcheck/linkparse.py

  * removed Gilman news draft
    Type: feature
    Removed: draft-gilman-news-url-00.txt


1.12.1 (release 21.2.2004)
  * raise IncompleteRead instead of ValueError on malformed chunked
    HTTP data
    Changed: linkcheck/httplib2.py
  * catch errors earlier in recursion check
    Changed: linkcheck/UrlData.py
  * quote url and parent url in log output
    Changed: linkcheck/log/*.py
    Added: linkcheck/url.py

1.12.0 (release 31.1.2004)
  * added LRU.setdefault function
    Changed: linkcheck/LRU.py
    Closes: SF bug 885916
  * Added Mac OS X as supported platform (version 10.3 is known to work)
    Changed: README, INSTALL
  * HTML parser objects are now subclassable and collectable by the cyclic
    garbage collector
    Changed: linkcheck/parser/htmlparse.y
  * made some minor parser fixes for attribute scanning and JavaScript
    Changed: linkcheck/parser/htmllex.l
  * include the optcomplete module for bash autocompletion
    Added: linkcheck/optcomplete.py, linkcheck-completion
    Changed: MANIFEST.in, setup.py
  * print out nicer error message for unknown host names
    Changed: linkcheck/UrlData.py
  * added new logger type "none" printing out nothing which is handy for
    cron scripts.
    Changed: linkchecker, linkcheck/Config.py, linkcheck/log/__init__.py
    Added: linkcheck/log/NoneLogger.py
  * the -F file output option disables console output now
    Changed: linkchecker
  * added an example cron script
    Added: linkcheck-cron.sh
    Changed: MANIFEST.in, setup.py
  * only warn about missing anchor support servers when the url has
    actually an anchor
    Changed: linkcheck/HttpUrlData.py
  * always fall back to HTTP GET request when HEAD gave an error to
    cope with servers not supporting HEAD requests
    Changed: linkcheck/HttpUrlData.py, FAQ

1.10.3 (release 10.1.2004)
  * use the optparser module for command line parsing
    Changed: linkchecker, po/*.po
  * use Set() instead of hashmap
    Changed: linkcheck/Config.py
  * fix mime-type checking to allow parsing of .css stylesheets
    Changed: linkcheck/HttpUrlData.py
  * honor HTML meta tags for robots, ie.
    <meta name="ROBOTS" content="NOFOLLOW">
    Changed: linkcheck/UrlData.py, linkcheck/linkparse.py
  * much less aggressive thread acquiring, this fixes the 100% CPU
    usage from the previous version
    Changed: linkcheck/Threader.py

1.10.2 (release 3.1.2004)
  * fixed CGI safe_url pattern, it was too strict
    Changed: linkcheck/lc_cgi.py
  * replace backticks with repr() or %r
    Changed: all .py files containing backticks, and po/*.po
  * make windows DNS nameserver parsing more robust
    Changed: linkcheck/DNS/Base.py
    Closes: SF bugs 863227,864383
  * only cache used data, not the whole url object
    Changed: linkcheck/{Http,}UrlData.py
  * limit cached data
    Changed: linkcheck/{UrlData,Config}.py
    Added: linkcheck/LRU.py
    Closes: SF bug 864516
  * use dummy_threading module and get rid of the _NoThreads
    functions
    Changed: linkchecker, linkcheck/{Config,Threader}.py,
      test/test_*.py
  * set default connection timeout to 60 seconds
    Changed: linkcheck/__init__.py
  * new option --status print regular messages about number of
    checked urls and urls still to check
    Changed: linkchecker, linkcheck/{__init__,Config}.py

1.10.1 (release 19.12.2003)
  * added Mandrake .spec file from Chris Green <cmg@dok.org>
    Added: linkchecker.spec
    Changed: MANIFEST.in
  * print last-modified date for http and https links in infos
    Changed: linkcheck/HttpUrlData.py
  * add detailed installation instructions for Windows
    Changed: INSTALL
    Closes: SF bug 857748
  * updated the DNS nameserver config parse routines
    Changed: linkcheck/DNS/Base.py
    Added: linkcheck/DNS/winreg.py
    Removed: linkcheck/DNS/win32dns.py
  * fix https support test
    Changed: linkcheck/HttpUrlData.py

1.10.0 (released 7.12.2003)
  * catch httplib errors in robotparser
    Changed: linkcheck/robotparser2.py
    Closes: SF bug 836864
  * - infinite recursion option with negative value works now
    - initialize self.urlparts to avoid crash when reading cached http
      urls
    - with --strict option do not add any automatic filters if the user
      gave his own on the command line
    Changed: linkcheck/UrlData.py

1.9.5 (released 31.10.2003)
  * Add Zope to servers with broken HEAD support, adjusted the FAQ
    Changed: linkcheck/HttpUrlData.py, FAQ
    Closes: SF bug 833419
  * Disable psyco usage, it is causing infinite loops (this is a known
    issue with psyco); and it is disabling ctrl-c interrupts (this
    is also a known issue in psyco)
    Changed: linkchecker
  * use internal debug logger
    Changed: linkcheck/robotparser2.py
  * do not hardcode Accept-Encoding header in HTTP request
    Added: linkcheck/httplib2.py
    Changed: linkcheck/robotparser2.py

1.9.4 (released 22.10.2003)
  * parse CSS stylesheet files and check included urls, for example
    background images
    Changed: linkcheck/{File,Http,Ftp,}UrlData.py, linkcheck/linkparser.py
  * try to use psyco for the commandline linkchecker script
    Changed: linkchecker
  * when decompression of compressed HTML pages fails, assume the page
    is not compressed
    Changed: linkcheck/{robotparser2,HttpUrlData}.py

1.9.3 (released 16.10.2003)
  * re-added an updated robot parser which uses urllib2 and can decode
    compressed transfer encodings.
    Added: linkcheck/robotparser2.py
  * more restrictive url validity checking when running in CGI mode
    Changed: linkcheck/lc_cgi.py
  * accept more Windows path specifications, like
    file://C:\Dokume~1\test.html
    Changed: linkcheck/FileUrlData.py

1.9.2
  * parser fixes:
    - do not #include <stdint.h>, fixes build on some FreeBSD, Windows
      and Solaris/SunOS platforms
    - ignore first leading invalid backslash in a=\"b\" attributes
    Changed: linkcheck/parser/htmllex.{l,c}
  * add full script path to linkchecker on windows systems
    Changed: linkchecker.bat
  * fix generation of Linkchecker_Readme.txt under windows systems
    Changed: setup.py

1.9.1
  * add documentation how to change the default C compiler
    Changed: INSTALL
  * fixed blacklist logging
    Changed: linkcheck/log/BlacklistLogger.py
  * removed unused imports
    Changed: linkcheck/*.py
  * parser fixes:
    - fixed parsing of end tags with trailing garbage
    - fixed parsing of script single comment lines
    Changed: linkcheck/parser/htmllex.l

1.9.0
  * Require Python 2.3
    - removed timeoutsocket.py and robotparser.py, using upstream
    - use True/False for boolean values
    - use csv module
    - use new-style classes
    Closes: SF bug 784977
    Changed: a lot
  * update po makefiles and tools
    Changed po/*
  * start CGI output immediately
    Changed: lc.cgi, lc.fcgi, lc.sz_fcgi, linkcheck/lc_cgi.py
    Closes: SF bug 784331

1.8.22
  * allow colons in HTML attribute names, used for namespaces
    Changed: linkcheck/parser/htmllex.l
  * fix match of intern patterns with --denyallow enabled
    Changed: linkcheck/UrlData.py
  * s/intern/internal/ and s/extern/external/ in the documentation
    Changed: linkchecker, linkchecker.1, FAQ
  * rename column "column" to "col" in SQL output, since "column" is
    a reserved keyword. Thanks Garvin Hicking for the hint.
    Changed: linkcheck/log/SQLLogger.py, create.sql
  * handle HTTP redirects to a non-http url
    Changed: linkcheck/{Http,}UrlData.py
    Closes: SF bug 784372

1.8.21
  * detect recursive redirections; the maximum of five redirections is
    still there though
  * after every HTTP 301 or 302 redirection, check the URL cache again
    Closes: SF bug 776851
  * put all HTTP 301 redirection answers also in the url cache as
    aliases of the original url. this could mess up some redirection
    warnings (ie warn about redirection when there is none), but it is
    more network efficient.

1.8.20
  * fix setting of domain in set_intern_url
    Changed: linkcheck/UrlData.py
  * - parse JS strings and comments
    - accept "<!- " as comment begin
    Changed: linkcheck/parser/htmlex.l
    Closes: SF bug 768661
  * quote url before submitting the request, the previous map() call
    was useless. Thanks Toby Dickenson for the patch.
    Changed: linkcheck/HttpUrlData.py
    Closes: SF bug 776416

1.8.19
  * add scheme colon in set_intern_url
    Changed: linkcheck/UrlData.py
  * fix threading option -t
    Changed: linkchecker, linkcheck/Config.py
  * do not try to get content of urls that have no content (eg mail)
    Closes: SF bug 765016
    Changed: linkcheck/{Mailto,Nntp,Telnet,}UrlData.py
  * added robots.txt FAQ, updated links
    Removed: norobots-rfc.html
    Changed: FAQ, WONTDO, TODO
  * add iso-8859-1 coding line to all .py files
    Changed: *.py
  * Correctly quote the HTML output
    Changed: linkcheck/log/HtmlLogger.py

1.8.18
  * fix option error messages for invalid integer arguments
    Changed files: linkchecker
  * enable infinite recursion with a negative -r value
    Changed files: linkcheck/{UrlData,Config}.py, linkchecker,
      linkchecker.1
  * if -s is given, add some link patterns to urls given on the
    command line automatically:
    for local files, add -i "^file:". For http and ftp urls, add
    the domain name -i "<domain>".
    Changed files: linkcheck/UrlData.py, linkchecker

1.8.17
  * fix parsing of missing end tag in "</a <a b=c>"
    Changed files: linkcheck/parser/htmllex.l
  * fix entity resolving in parsed html links
    Closes: SF bug #749543
    Changed files: linkcheck/StringUtil.py

1.8.16
  * also look at id attributes on anchor check
    (Closes SF Bug #741131)
    Changed files: linkcheck/{linkparser,UrlData}.py
  * minor parser cleanups
    Changed files: linkcheck/parser/*

1.8.15
  * Fix compile errors with C variable declarations in HTML parser.
    Thanks to Fazal Majid <fazal@majid.fm>
    Changed files: linkcheck/parser/htmlparse.[yc]

1.8.14
  * fix old bug in redirects not using the full url. This resulted in
    errors like (-2, "Name or service not known")
    Changed files: linkcheck/HttpUrlData.py
    Closes: SF Bug #729007
  * only remove anchors on IIS servers (other servers are doing quite
    well with anchors... can you spell A-p-a-c-h-e ?)
    Changed files: linkcheck/{HttpUrlData, UrlData}.py
  * Parser changes:
    - correctly propagate and display parsing errors
    - really cope with missing ">" end tags
    Changed files: linkcheck/parser/html{lex.l, parse.y},
      linkcheck/linkparse.py, linkcheck/UrlData.py
  * quote urls before a request
    Changed files: linkcheck/HttpUrlData.py

1.8.13
  * fix typo in manpage
    Changed files: linkchecker.1
  * remove anchor from HEAD and GET requests
    Changed files: linkcheck/{HttpUrlData, UrlData}.py

1.8.12
  * convert urlparts to list also on redirect
    Changed files: linkcheck/HttpUrlData.py

1.8.11
  * catch httplib.error exceptions
    Changed files: linkcheck/HttpUrlData.py
  * override interactive password question in robotparser.py
    Changed files: linkcheck/robotparser.py
  * switch to urllib2.py as default url connect.
    Changed files: linkcheck/UrlData.py
  * recompile html parser with flex 2.5.31
    Changed files: linkcheck/parser/{htmllex.c,Makefile}

1.8.10
  * new option --no-anchor-caching
    Changed files: linkchecker, linkcheck/{Config.py, UrlData.py}, FAQ
  * quote empty attribute arguments
    Changed files: linkcheck/parser/htmllex.[lc]

1.8.9
  * recompile with bison 1.875a
    Changed files: linkcheck/parser/htmlparse.[ch]
  * remove stpcpy declaration, fixes compile error on RedHat 7.x
    Changed files: linkcheck/parser/htmlsax.h
  * clarify keyboard interrupt warning to wait for active connections
    to finish
    Changed files: linkcheck/__init__.py
  * resolve &#XXX; number entity references
    Changed files: linkcheck/{StringUtil.py,linkname.py}

1.8.8
  * All amazon servers block HEAD requests with timeouts. Use GET as
    a workaround, but issue a warning.
    Changed files: linkcheck/HttpUrlData.py
  * restrict CGI access to localhost per default
    Changed files: lc.cgi, lc.fcgi, lc.sz_fcgi, linkcheck/lc_cgi.py

1.8.7
  * #define YY_NO_UNISTD_H on Windows systems, fixes build error with
    Visual Studio compiler
    Changed files: setup.py
  * use python2.2 headers for parser compile, not 2.1.
    Changed files: linkcheck/parser/Makefile

1.8.6
  * include a fixed robotparser.py (from Python 2.2 CVS maint branch)

1.8.5
  * fix config.warn to warn
    Changed files: linkcheck/__init.py
  * parser changes:
    o recognise "<! -- -->" HTML comments (seen at Eonline)
    o recognise "<! !>" HTML comments (seen at www.nba.com)
    o rebuild with flex 2.5.27
    Changed files: linkcheck/parser/htmllex.[lc]
  * added another url exclusion example to the FAQ
    numerate questions and answers
    Changed files: FAQ
  * fix linkchecker exceptions
    Changed files: linkcheck/{Ftp,Mailto,Nntp,Telnet,}UrlData.py,
      linkcheck/__init__.py

1.8.4
  * Improve error message for failing htmlsax module import
    Changed files: linkcheck/parser/htmllib.py
  * Regenerate parser with new bison 1.875
    Changed files: linkcheck/parser/htmlparser.c
  * Some CVS files were not the same as their local counterpart.
    Something went wrong. Anyway, I re-committed them.
    Changed files: a lot .py files
 
1.8.3
  * add missing imports for StringUtil in log classes, defer i18n of log
    field names (used for CGI scripts)
    Changed files: linkcheck/log/*.py
  * fixed wrong debug level comparison from > to >=
    Changed files: linkcheck/Config.py
  * JavaScript checks in the CGI scripts
    Changed files: lconline/lc_cgi.html.*
    Added files: lconline/check.js
  * Updated documentation with a link restriction example
    Changed files: linkchecker, linkchecker.1, FAQ
  * Updated po/pygettext.py to version 1.5, cleaned up some gettext
    usages.
  * updated i18n
    Added files: linkcheck/i18n.py
    Changed files: all .py files using i18n
  * Recognise "<! --" HTML comments
    Changed files: linkcheck/parser/htmllex.l
  * -a anchor option implies -w because anchor errors are always warnings
    Changed files: linkchecker
  * added AnsiColors.py and debug.py to split out some functions
    Changed files: a lot .py files using these things
  * use yy_size_t for parser alloc definitions, fixes build errors on 64bit
    architectures
    Changed files: linkcheck/parser/htmllex.l

1.8.2
  * - ignore invalid html attribute characters
    - ignore trailing garbage on html end tags
    - fixed debugging code with flex
    - use flex memory management interface
    - use only double quotes for attribute quoting
    - check quoting of all attributes
    Changed files: linkcheck/parser/htmllex.l
  * build parser with flex 2.5.25
    Changed files: linkcheck/parser/{Makefile, htmllex.c}
  * put shared code of cgi scripts in lc_cgi.py
    Changed files: lc.cgi, lc.fcgi, lc.sz_fcgi, linkcheck/lc_cgi.py
  * put some linebreaks and target="top" into HTML output
    Changed files: linkcheck/logging/HtmlLogger.py
  * add translated cgi files
    Changed files: setup.py, MANIFEST.in, debian/rules
    Added files: lconline/*.{de,en}
    Removed files: lconline/{leer.html,lc_cgi.html}

1.8.1
  * Add missing () to function call in proxy handling code
    Changed files: FtpUrlData.py
  * Use urlparse.url(un)split instead of urlparse.url(un)parse
    Changed files: FtpUrlData.py, UrlData.py, HttpUrlData.py,
      FileUrlData.py
  * Print size information if its available
    Changed files: FtpUrlData.py, UrlData.py, HttpUrlData.py
  * Add --warning-size-bytes option to print warning if content size
    exceeds the given byte limit
    Changed files: FtpUrlData.py, HttpUrlData.py, linkchecker, Config.py,
      linkchecker.1
  * Updated translations
    Changed files: po/linkchecker.pot, po/*.po
  * Parse supported file types for ftp links
    Changed files: FtpUrlData.py, FileUrlData.py, UrlData.py

1.8.0
  * Require Python >= 2.2.1, remove httplib.
    Changed files: setup.py, INSTALL, linkchecker
  * Add again python-dns, the Debian package maintainer is unresponsive
    Added files: linkcheck/DNS/*.py
    Changed files: INSTALL, setup.py
  * You must now use named constants for ANSII color codes
    Changed files: linkcheckerrc, linkcheck/log/ColoredLogger.py
  * Release RedHat 8.0 rpm packages.
    Changed files: setup.py, MANIFEST.in
  * remove --robots-txt from manpage, fix HTZP->HTTP typo
    Changed files: linkchecker.1

1.7.1
  * Fix memory leak in HTML parser flushing error path
    Changed files: htmlparse.y
  * add custom line and column tracking in parser
    Changed files: htmllex.l, htmlparse.y, htmlsax.h, htmllib.py
  * Use column tracking in urldata classes
    Changed files: UrlData.py, FileUrlData,py, FtpUrlData.py,
     HostCheckingUrlData.py
  * Use column tracking in logger classes
    Changed files: StandardLogger.py CVSLogger.py, ColoredLogger.py,
      HtmlLogger.py, SqlLogger.py

1.7.0
  * Added new HTML parser written in C as a Python extension module.
    It is faster and it is more fault tolerant.
    Of course, this means I cannot provide .exe installers any more
    since the distutils dont provide cross-compilation.

1.6.7
  * Removed check for <applet> tags codebase attribute, but honor it
    when checking applet links
  * Handle <applet> tags archive attribute as a comma separated list
    Closes: SF bug #636802
  * Fix a nasty bug in tag searching, which ignored tags with more
    than one link attribute in it.
  * Fix concatenation with relative base urls by first joining the
    parent url.
  * New commandline option --profile to write profile data.
  * Add httplib.py from Python CVS 2.1 maintenance branch, which has the
    skip_host keyword argument I am using now.

1.6.6
  * Use the new HTTPConnection/HTTPResponse interface of httplib
    Closes: SF bug #634679
    Changed files: linkcheck/HTTPUrlData.py, linkcheck/HTTPSUrlData.py
  * Updated the ftp online test
    Changed files: test/output/test_ftp

1.6.5
  * Catch the maximum recursion limit error while parsing links and
    print an error message instead of bailing out.
    Changed files: linkcheck/UrlData.py
  * Fixed Ctrl-C only interrupting one single thread, not the whole
    program.
    Changed files: linkcheck/UrlData.py, linkcheck/__init__.py
  * HTML syntax cleanup and relative cgi form url for the cgi scripts
    Changed files: lconline/*.html

1.6.4
  * Support for ftp proxies
    Changed files: linkcheck/FtpUrlData.py, linkcheck/HttpUrlData.py
    Added files: linkcheck/ProxyUrlData.py
  * Updated german translation

1.6.3:
  * Generate md5sum checksums for distributed files
    Changed files: Makefile
  * use "startswith" string method instead of a regex
    Changed files: linkchecker, linkcheck/UrlData.py
  * Add a note about supported languages, updated the documentation.
    Changed files: README, linkchecker, FAQ
  * Remove --robots-txt option from documentation, it is per default
    enabled and you cannot disable it from the command line.
    Changed files: linkchecker, po/*.po
  * fix --extern argument creation
    Changed files: linkchecker, linkcheck/UrlData.py
  * Print help if PyDNS module is not installed
    Changed files: linkcheck/UrlData.py
  * Print information if a proxy was used.
    Changed files: linkcheck/HttpUrlData.py
  * Updated german documentation
    Changed files: po/de.po
  * Oops, an FTP proxy is not used. Will make it in the next release.
    Changed files: linkcheck/FtpUrlData.py
  * Default socket timeout is now 30 seconds (10 was too short)

1.6.2:
  * Warn about unknown Content-Encodings. Dont parse HTML in this case.
  * Support deflate content encoding (snatched from Debians reportbug)
  * Add appropriate Accept-Encoding header to HTTP request.
  * Updated german translations

1.6.1:
  * FileUrlData.py: remove searching for links in text files, this is
    error prone. Just handle *.html and Opera Bookmarks.
  * Make separate ChangeLog from debian/changelog. For previous
    changes, see debian/changelog.
  * Default socket timeout is now 10 seconds
  * updated linkcheck/timeoutsocket.py to newest version
  * updated README and INSTALL
  * s/User-agent/User-Agent/, use same case as other browsers
