CGI::Session Change Log
=====================================================================
4.38 - Friday, October 31, 2008
* INTERNAL: Rename SimpleObjectClass to CGI::Session::Test::SimpleObjectClass
to avoid namespace ownership issue (Mark Stosberg).
* INTERNAL: We now list CGI.pm 3.26 or greater as a dependency. You are still
welcome to use other query objects, but this version of CGI.pm
fixes a bug in the strictness of HTTP expiration times, which
Safari in particular is sensitive to. So, if you are using
CGI.pm, you should upgrade to at least this version. RT#34216,
thanks to Astar, Michael Hampton, Ron Savage and Mark Stosberg.
* INTERNAL: return explicit values in _set_status and _unset_status
(RT#39201, Mario Domgoergen, Mark Stosberg)
* FIX: RT#37877: The storable serializer wasn't properly inheriting the 'errstr'
method. This could have resulted an error like:
"Can't locate errstr via package "CGI::Session::Serialize::storable"
Thanks to Michael Greenish, Mark Stosberg.
* FIX: RT#40405 reported a case where the default serializer would have a problem after the user
set a parameter's value to undef, in certain circumstances.
A test file was kindly provided by cowomally[...]nullium.net.
The fix was spelled out by Matt LeBlanc
* FIX: RT#39679 pointed out a simplification in method remove() in CGI::Session::Driver::file.
By calling _file() instead of duplicating code, we get the benefit of extra error checking.
Thanx to Sergiy Borodych for noticing this
* FIX: Stop using the return value of delete() in t/find.t. This means that when the patch
provided in RT#37752 is applied, t/find.t will not start failing
4.37 - Wednesday, October 22, 2008
* INTERNAL: Patch Makefile.PL and Build.PL to request that SimpleObjectClass not be indexed.
4.36 - Friday, September 12, 2008
* FIX: The sample code for find() had 2 errors in it:
o It assumed delete() returned a meaningful value, which it doesn't
o It did not follow the call to delete() with a (recommended) call to flush()
o Thanks to Mario Domgoergen for the report, RT#39201
4.35 - Tuesday, July 15, 2008
* FIX: Fix the bug fix for error propagation from 4.34, RT#37628
* FIX: Reset errors when creating a new object. Previously, errors from
a previous object could show up in the current object.
* INTERNAL - Patch Build.PL and Makefile.PL to run Module::Metadata::Changes's ini.report.pl,
if available, to regenerate Changelog.ini from this file (Changes), when creating a makefile.
4.34 - Sunday, July 13, 2008
* SECURITY: Patch CGI::Session::Driver::file to stop \ and / characters being used in
session ids and hence in file names. These characters, possibly combined with '..',
could have been used to access files outside the designated session file directory.
Reported by TAN Chew Keong of vuln.sg.
* FIX: Patch CGI::Session to propagate error upwards when _load_pluggables() fails.
See RT#37628 and http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=490198.
* INTERNAL: Ship a machine-readable version of this file under the name Changelog.ini.
The latter file is generated by ini.report.pl, which is shipped with Module::Metadata::Changes.
The reason Changelog.ini does not contain a separate section for each version in this file
is that some of the versions documented below have no datestamp, and ini.report.pl does not create
fake datestamps.
4.33 - Monday, July 7, 2008
* FIX: Patch CGI::Session::Driver::mysql to replace 'REPLACE INTO ...' with
'INSERT INTO ... ON DUPLICATE KEY UPDATE ...'. See RT#37069.
Thanks to Steve Kirkup for the patch. I (Ron) installed MySQL V 5.0.51a for testing.
Note: http://dev.mysql.com/doc/refman/5.0/en/releasenotes-cs-5-0-45.html and similar docs
list various MySQL errors fixed recently for the above new syntax. Also, the new version
is now much more like the Postgres code, which is another reason it has been adopted.
* FIX: t/mysql.t used to test setting the global variable $CGI::Session::MySQL::TABLE_NAME.
The test for this (in t/mysql.t) was introduced in V 4.00_09.
However, since V 4.29_1, changes to CGI::Session::Driver's new() method mean
this way of setting the session table's name no longer works, and so the variable
$CGI::Session::MySQL::TABLE_NAME is now not used. Hence it has been removed.
Code in CGI::Session::Driver::DBI used to set $class::TABLE_NAME for all database drivers.
This code has also been removed. Moral: Don't use global variables.
Call $session = CGI::Session -> new(..., ..., (TableName => 'new_name'}) or,
after creating the object, call $session -> table_name('new_name').
To retrieve the name, call $name = $session -> table_name().
4.32 - Tuesday, June 17, 2008
* FIX: Packaging of 4.31 release was botched.
4.31 - Tuesday, June 10, 2008
* FIX: Patch CGI::Session::Driver::DBI to check that the DBI handle still exists before trying
to ping it. This handles the case where the DBI object is destroyed before the session object.
See RT#35925.
* FIX: Patch CGI::Session::Driver::DBI's remove() which still hard-coded the column name 'id' instead
of using the new feature which allows the user to specify the name of the column. See RT#36235.
* FIX: Patch POD yet again to emphasize that an explicit call to destroy() should be followed by
explicit call to flush(), in particular in the case where the program is not exiting and
hence auto-flushing is not activated. Sections patched are 'A Warning about Auto-flushing'
and the docs for delete(). See RT#34668.
4.30 - Friday, April 25, 2008
* FIX: Patch POD for CGI::Session in various places, to emphasize even more that auto-flushing is
unreliable, and that flush() should always be called explicitly before the program exits.
The changes are a new section just after SYNOPSIS and DESCRIPTION, and the PODs for flush(),
and delete(). See RT#17299 and RT#34668
* NEW: Add t/new_with_undef.t and t/load_with_undef.t to explicitly demonstrate the effects of
calling new() and load() with various types of undefined or fake parameters. See RT#34668
* FIX: Patch POD for new() and load() to clarify the result of calling these with undef, or with
an initialized CGI object with an undefined or fake CGISESSID. See RT#34668.
Specifically: You are strongly advised to run the old-fashioned
'make test TEST_FILES=t/new_with_undef.t TEST_VERBOSE=1' or the new-fangled
'prove -v t/new_with_undef.t', for both new*.t and load*.t, and examine the output
* FIX: Patch POD in various tiny ways to improve the grammar
4.29_2 - Thursday, March 27, 2008
* FIX: stop ExtUtils::MakeMaker trying to create Build.PL (Ron Savage)
* FIX: Disable trying to use utf8 in tests. (Ron Savage) Ref RT#21981, RT#28516
4.29_1 - Saturday, March 15, 2008
Special Thanks to Ron Savage who did the bulk of the work to put this release together.
* FIX: Patch CGI::Session to fix RT#29138 (Patch by Barry Friedman)
* NEW: Add a note to CGI::Session's POD referring to utf8 problems, and include references
to RT#21981 (Reported by erwan) and RT#28516 (Reported by jasoncrowther)
* FIX: Patch CGI::Session::Driver::DBI.pm to fix RT#24601 (Patch by latypoff)
* FIX: Patch CGI::Session::Driver::DBI.pm to fix RT#24355 (Reported by fenlisesi, patch by Ron Savage)
* NEW: Add t/bug24285.t to ensure session data files are created properly when the user specifies a
directory other than /tmp (Reported by William Pearson RT#24285, patch by Ron Savage)
* FIX: Patch t/ip_matches.t and t/bug21592.t to remove test files left in /tmp, to fix RT#29969
(Reported by ANDK, patch by Ron Savage)
* FIX: Patch POD for CGI::Session::Driver::file to clarify how to use the option to change the
file name pattern used to created session files (Report by appleaday RT#33635,
patch by Ron Savage)
* FIX: Patch CGI::Session::Driver::sqlite to add sub DESTROY to fix RT#32932
(Patch by Alexander Batyrshin, corrected by Ron Savage)
* FIX: Remove CGI::Session::Seralize::json and t/g4_dbfile_json.t until such time as this code
can be made to work reliably. Both JSON::Syck and JSON::XS have been tried, and in both
cases t/g4_dbfile_json.t dies horribly (but differently). Patch POD for CGI::Session to
remove references to JSON. RT#25325 (Reported by bkw, patch by Ron Savage)
* NEW: Patch CGI::Session's POD and load() to allow the session/cookie name default of CGISESSID
to be overridden. (Patch by Lee Carmichael RT#33437, reformatted by Ron Savage). Lee has
also patched t/name.t to test the new functionality
* NEW: Split CGI::Session::Serialize::yaml out into its own distro. Get it hot from CPAN!
* NEW: Add Build.PL for Module::Build users. This also requires adding PL_FILES => {}
to Makefile.PL to beat ExtUtils::MakeMaker over the head, otherwise it executes
'perl Build.PL Build'
* NEW: Support specification of both the id column name and the a_session column name in the
sessions table, by extending the options acceptable in CGI::Session->new(..,..,{here}).
Allow: {TableName => 'session', IdColName => 'my_id', DataColName => 'my_data'}.
Default: {TableName => 'sessions', IdColName => 'id', DataColName => 'a_session'}.
Allow any 1, 2 or 3 of these options. Missing keys default as specified.
(Patch by Chris RT#2224. Implemented differently by Ron Savage). Supported drivers:
o MySQL (native to CGI::Session)
o ODBC (separate distro, CGI::Session::Driver::odbc V 1.01)
o Oracle (separate distro, CGI::Session::Driver::oracle V 1.01)
o Postgres (native)
o SQLite (native)
4.20 - Monday, December 4, 2006
* INTERNAL: No Changes since 4.20_1. Declaring stable.
4.20_1 - Friday, November 24, 2006
* FIX: -ip_match now works even when it's not the last import item. (RT#21779)
* FIX: In the PostgreSQL driver, a race condition is when storing is now worked around. (Mark Stosberg)
* FIX: Added important clarification and example to MySQL driver docs that the session column
needs to be defined as a primary key to avoid duplicate sessions. (Justin Simoni, Mark Stosberg)
* FIX: The default serializer now works correctly with certain data structures. (RT#?) (Matt LeBlanc)
* FIX: A documentation bug in find() was fixed (Matt LeBlanc)
* FIX: Documented how to declare a database handle to be used on demand, which was introduced
in 4.04. (Mark Stosberg)
* FIX: Connections made with SQLite now disconnect only when appropriate, instead of always.
This addresses a symptom seen as "attempt to prepare on inactive database handle"
(Jaldhar Vyas, Sherzod, Mark Stosberg)
* FIX: Args to the constructor for CGI::Session and the drivers are now always shallow
copied rather than used directly, to prevent modification.
(RT#21952, Franck Porcher, Sherzod, Mark Stosberg)
* FIX: The documentation for expire($param, $time) was made more explicit
(pjf, Mark Stosberg)
* NEW: Added recommended use of flush() to the Synopsis (Michael Renner, RT#22333)
* NEW: Added links to Japanese translations of the documentation (Makio Tsukamoto)
http://digit.que.ne.jp/work/index.cgi?Perldoc/ja
* INTERNAL: Update test to workaround YAML versions less than 0.58. (Matt LeBlanc)
* INTERNAL: param() code was refactored for clarity (Mark Stosberg, Ali ISIK, RT#21782)
* INTERNAL: new() and load() were refactored (Ali Isik)
* INTERNAL: renamed some environment variables used for testing (Ron Savage)
* INTERNAL: Multi key-value syntax of param() now always returns number of keys
successfully processed, 0 if no key/values were processed.
4.14 - Sunday, June 11, 2006
* NEW: The find() command now has better documentation. (Ron Savage, Matt LeBlanc)
* FIX: find() no longer changes the access or modified times (RT#18442) (Matt LeBlanc)
* FIX: param() called with two parameters now returns the value set, if any (RT#18912) (Matt LeBlanc)
* FIX: driver, serializer, and id generator names are now untainted (RT#18873) (Matt LeBlanc)
* INTERNAL: automatic flushing has been documented to be unreliable, although
it was recommended in the past. Automatic flushing can be affected adversely
in persistent environments and in some cases by third party software. There are
also some cases in which flushing happened automatically in 3.x, but quit working
with 4.x. See these tickets for details.
http://rt.cpan.org/Ticket/Display.html?id=17541
http://rt.cpan.org/Ticket/Display.html?id=17299
4.13 - Wednesday, April 12, 2006
* FIX: Applied patch to fix cookie method (RT#18493,Nobuaki ITO)
* FIX: Berkeley DB 1.x exhibits a bug when used in conjunction with O_NOFOLLOW. Because of this,
we've removed it from the db_file driver. It will still attempt to stop symlinks but the
open itself has dropped the flag. (Matt LeBlanc)
* FIX: json and yaml db_file tests now check for the presence of DB_File. (Matt LeBlanc)
4.12 - Friday, April 7, 2006
* SECURITY: Fix possible SQL injection attack. (RT#18578, DMUEY)
4.11 - Friday, March 31, 2006
* FIX: Since 4.10, using name() as a class method was broken. This has
been fixed, and regression tests for both uses have been added. (Matt LeBlanc)
4.10 - Tuesday, March 28, 2006
* SECURITY: Hopefully this settles all of the problems with symlinks. Both the file
and db_file drivers now use O_NOFOLLOW with open when the file should exist and
O_EXCL|O_CREAT when creating the file. Tests added for symlinks. (Matt LeBlanc)
* SECURITY: sqlite driver no longer attempts to use /tmp/sessions.sqlt when no
Handle or DataSource is specified. This was a mistake from a security standpoint
as anyone on the machine would then be able to create and therefore insert data
into your sessions. (Matt LeBlanc)
* NEW: name is now an instance method (RT#17979) (Matt LeBlanc)
4.09 - Friday, March 16th, 2006
* SECURITY: Applying security patch from: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=356555 (Julien Danjou)
4.08 - Thursday, March 15th, 2006
* FIX: DESTROY was sometimes wiping out exception handling. RT#18183, Matt LeBlanc.
* SECURITY: Resolve some issues in: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=356555
- db_file and file now check for symlinks either explicitly or by using O_EXCL on sysopen
- file creation umask defaults to 660
* NEW: db_file and file drivers now accepts a UMask option. (Matt LeBlanc)
* INTERNAL: test suite clean up (Tyler MacDonald)
4.07 - Thursday, March 9th, 2006
* INTERNAL: MANIFEST update to fix release.
4.06 - Wednesday, March 3rd, 2006
* INTERNAL: MANIFEST update to fix release.
4.06 - Wednesday, March 8th, 2006
* FIX: some stray warnings when flushing: "Use of uninitialized value in numeric eq (==)" (RT#14603)
* NEW: JSON and YAML serializers (Tyler MacDonald)
* INTERNAL: CGI::Session::Test::Default accepts a "skip" argument,
listing tests that should be skipped. (Tyler)
4.05 - Friday, March 3rd, 2006
* FIX: Race condition fixed when writing to session files (RT#17949)
4.04 - Wednesday, March 01, 2006
* NEW: File driver now has option to disable flock (for those running
Win 9x, VMS, MacPerl, VOS and RISC OS). (Matt LeBlanc)
* FIX: If DBI driver was initialized using 'Handle', Driver::DBI::init()
returned false, and Driver::new() thought init faild and kept returning
undef. The problem was fixed by making sure Driver::DBI::init() returned
true. (Sherzod)
* Added .*cgisess.* to disclude cgisess.db, cgisess.id, and any session
files created in the t directory. (Matt LeBlanc)
* FIX: File driver now respects $CGI::Session::File::FileName for 3.9x
compatibility. (Matt LeBlanc)
* FIX: Default serializer now properly handles data structures that appear
more than once in the serialized data structure (before it'd result in data
structures that were equivalent but did not have the same address). (Matt LeBlanc)
* FIX: File driver now localizes the filehandle to avoid any possibility
of extended locking in persistent environments (Matt LeBlanc)
* FIX: File driver now locks the file when retrieving the session data (Matt LeBlanc)
* NEW: DBI Drivers now support a lazy loaded database handle. This is useful with the
CGI::Application plugin system. If the session is never used, the database handle may not
not need to be created. The syntax is to use a code ref:
Handle => sub {DBI->connect} (Mark Stosberg)
Finally, be aware that since 4.0 some people have reported problems with
the auto-flushing code. There may be an unresolved. You always call
flush() to be safe. Input or code contributions for the issue are
appreciated. Some related tickets include:
http://rt.cpan.org/Public/Bug/Display.html?id=14604
http://rt.cpan.org/Public/Bug/Display.html?id=16861
http://rt.cpan.org/Public/Bug/Display.html?id=17541
http://rt.cpan.org/Public/Bug/Display.html?id=17299
4.03 - Wednesday, October 05, 2005
* FIX: automatic flushing did not work if session object was global
* FIX: Default serializer can now serialize objects (Matt LeBlanc)
* INTERNAL: SQLite driver no longer needs MIME::Base64 for encoding (Matt LeBlanc)
4.02 - Friday, September 02, 2005
* FIX: remote_addr() was missing (RT #14414])
4.01 - Thursday, September 01, 2005
* FIX: Minor POD fix
4.00 - Wednesday, August 31, 2005
*** NOTE ***
The 4.0 release represents a major overhaul of the CGI::Session code base.
Care has been taken to be 100% compatible with applications developed with 3.x.
However, you are encouraged to run regression tests with your own applications
before using this in production.
* NEW: PostgreSQL driver enhanced to work better with binary serializers (Matt LeBlanc)
* FIX: update to un tainting in default serializer to make "-T" happy (Matt LeBlanc)
* FIX: CGI::Session (qw/-ip_match/), a 3.x feature, works again (Shawn Sorichetti)
* INTERNAL: Improved documentation shown during "make", which explains how to run
database-driven tests. (Mark Stosberg)
* FIX: to support binary serializers SQLite driver uses MIME::Base64 (Sherzod Ruzmetov)
4.00_09 - Thursday, July 21, 2005
* CHANGE: Starting with 4.0, it will no longer work to use the syntax of
CGI::Session::DriverName(). This hasn't been a documented API since CGI::Session 2.94,
released in August, 2002.
* FIX: documented etime(), which was present in 3.x (Mark Stosberg)
* FIX: Added code, test and docs to make $CGI::Session::File::FileName work,
for 3.x compatibility. (Mark Stosberg)
* FIX: Providing an expire time like "-10" now works (Mark Stosberg)
* FIX: Restored close() method, for 3.x compatibility. (Mark Stosberg)
* FIX: Make ->clear('email') work, for 3.95 compatibility (Mark Stosberg)
* FIX: Added back is_new() for compatibility with 3.95. (Mark Stosberg)
* FIX: Support for CGI::Simple is confirmed, resolving RT#6141 (Mark Stosberg)
* FIX: Add code and tests for $CGI::Session::MySQL::TABLE_NAME, which worked in 3.x (Mark Stosberg)
* DOCS: CGI::Session now has a public Subversion repository, thanks to Jason Crome.
See the bottom of the CGI::Session docs for details.
4.00_08 - Tuesday, March 15, 2005
* FIX: Changes made in 4.00_07 rolled back
4.00_07 - Sunday, March 13, 2005
* FIX: overloaded objects are now stored properly
4.00_06 - Thursday, February 24, 2005
* FIX (?): a test script was failing on Win32
* FIX: inaccurate error reporting in load()
4.00_05 - Tuesday, February 22, 2005
* FIX: case insensitivity was not enforced properly in CGI::Session::parse_dsn()
4.00_04 - Wednesday, February 16, 2005
* FIX: Minor fix in tests suits and error-checking routines of
serializers and id-generators
4.00_03 - Friday, February 11, 2005
* NEW: CGI::Session::find() introduced
* NEW: traverse() introduced into drivers to support CGI::Session::find()
* DOCS: More complete driver specs documented
4.00_02 - Wednesday, February 09, 2005
* FIX: race conditions in Driver/file.pm pointed out by Martin Bartosch
4.00_01 - Wednesday, February 09, 2005
* NEW: load() - constructor method to prevent unnecessary session creations
* NEW: is_expired() - method to intercept expired sessions
* NEW: is_empty() - to intercept requests for un existing sessions
* NEW: more optimized source code
* NEW: updated and improved driver specs
* NEW: standard testing framework
* NEW: 'sqlite' driver
3.12
* cache() method introduced, which is normally used by library drivers to
cache certain value within the single process
* Apache::Session-like tie interface supported (EXPERIMENTAL!)
* trace() and tracemsg() methods added for debugging purposes
3.8
* Abbreviations in DSN parameters are supported via Text::Abbrev.
* Automatic api3 detection makes "-api3" switch obsolete
* Experimental "-frozen" switch added, but not yet functional.
* sync_param() utility function added
* header() replacement to CGI::header() added, which outputs
proper HTTP headers with session information
* Private data records have been documented.
* Bug in clear() kept failing if passed no arguments to be cleared.
3.x
* Ability to choose between serializers, drivers and id generators
while creating the session object. Supported via '-api3' switch.
* New serializers added: Storable, FreezeThaw in addition to Default.
* New ID generator added: Incr, which generates auto incrementing
id numbers, in addition to MD5
* "-ip_match" switch enabled for additional security
* Expire() method is fully functional
* Ability to expire certain session parameters
* Better documented drivers specifications
* Main documentation is split into two:
1) CGI::Session and 2) CGI::Session::Tutorial
* Bug in POD documentation is fixed (thanks to Graham Barr)
$Id: Changes 446 2008-11-01 03:01:15Z ron $