Revision history for Perl extension Net::FTPSSL.

Version - YYYY/MM/DD
0.23 2013/08/01 08:30:00
  - Added clarification to use of SSL_Client_Certificate option in pod text.
  - Added Bug Id 82094 to support tunneling through a proxy server via the
    ProxyArgs option.  Supports both proxy servers that require a password as
    well as those that don't.  This option requires Net::HTTPTunnel to use.
    But this new module isn't required if you don't use tunneling.
    (perl Makefile.PL will ask.)
  - Added optional proxy server questions if Net::HTTPTunnel is installed for
    the test scripts (t/*.t).  It warns you if it's missing during the test.
  - Fixed bug in login() so the call to help() only happens if the login is
  - Bug Id 84376 uncoverd bug in IO-Socket-SSL versions 1.79 to 1.85 when
    using Certificats for the Data Channel & getting a Man-In-The-Middle (MITM)
    warning that didn't apply.  So skip those releases.
  - Bug Id 85420 same MITM warning as the other bug, but this time without using
    Certificates in the command channel.  This required a code change in new()
    to fix.
  - Fixed masking issue when writing to the log file when the value that needed
    to be masked contained regex special chars.
  - Added clarification that any warnings printed below the 10-complex.t tests
    are just that.  Warnings, not errors preventing things from working.  They
    are collected so the developer can be notified about them for fixing in
    furture releases.

0.22 2012/05/21 08:30:00
  - Bug # 77072 requested to autodetect the need for using the PRET command.
    The old test in t/10-complex.t was different & doesn't interfere.  The docs
    in the POD now mark the Pret option as optional for cases where auto-detect
    doesn't work!
  - Fixed issue with ccc that the server sometimes responds to a successful NOOP
    twice, resulting in command/responses getting out of sync.  (Another hack)
  - Fixed quot() bug introduced with Bug # 74025 uc() fix in the last release.
    Accidently shifted off the arguments a 2nd time during the uc() call.
  - Fixed the return code in restart() to always be 1 instead of undef, since
    it never fails.  Also moved its description towards the end of the POD text.

0.21 2012/02/15 08:30:00
  - Fixed bad call to _croak_or_return () in _get_data_channel() when dealing
    with failed certificates for data connections.
  - Bug # 74547, a request to support ftp PRET command for distributed FTP data
    channel transfers.  New option to execute PRET before all PASV calls.
  - updated t/10-complex.t to dynamically check if PRET is needed.
  - Bug # 74035, as of Perl 5.12, uc($x) gives warning if $x is undef.  Added
    checks to avoid these warnings.  Did via [ uc($x||""); ]
  - Fixed both test cases to no longer send Port as undefined to get rid of
    warnings in log file. (Passing undef wasn't realistic, was just convenient.)
  - t/10-complex.t now explicity checks if any warnings are generated and does
    a request for feadback instead of as an error if any are found.  Should help
    considerably with debugging any code changes in the future.
  - Added new internal method _feat() to get additional commands to be
    recognized by supported().
  - Enhanced _help() in parsing it's response from the server & other significnt
    enhancements.  Major rewrite of this one method.
  - Above rewite uncovered bug in t/10-complex.t to fix in a supported() test.
  - Did some tweaking to the workings of the OverrideHELP option due to changes
    in _help() & supported().
  - Updated POD for supported() to describe changes to functionality.

0.20 2012/01/01 08:30:00
  - Fixed bug # 73115 where it looks like the response() had unexpected breaks
    in it.  Instead of returning "xyz message" the FTPS server was returning
    "x" on one line and "yz message" on another line.  And the code was
    treating "x" as the entire response (since there was no "-" after it.)
    Fixed so that if a response starts with a number that's less than 3
    digits with nothing else following it, to assume we hit this bug.
    The risk to this fix is that we might try to read past the response
    on some other server and cause things to hang.  So that's an issue to
    keep an eye out for after this release.
  - Fixed minor issue with "HELP" not supported on some boxes.  If this happens
    supported() will now always return false unless you use OverrideHELP.
  - Now prints the hash values chosen for new() to the logs.
  - Fixed so scrubbing out sensitive info from the logs is only done when debug
    is turned on instead of doing it all the time.
  - Reorganized the options to new() in the POD to put the less useful and
    special case options to new() last.
  - Other minor POD corrections and additions.

0.19 2011/09/26 08:30:00
  - If you use SSL_Advanced, it now writes to STDERR that you are using a
    depreciated feature.
  - Modified 10-complex.t & 20-certificate.t to honor the PERL_MM_USE_DEFAULT
    envionment variable for the smoke testers to always use defaults, which
    skips executing these 2 scripts.  BUG # 69982.
    (Same thing now happens for "make test < /dev/null" as well.)
    Couldn't use ExtUtils::MakeMaker::prompt() per request since "make test"
    wouldn't display the questions it was asking & it looked like it hung in
    interactive mode for normal users!  So based my solution on that function
    using diag() instead of print.
  - Fixed Typos in POD text Bug # 70438 & some misleading comments.
  - Updated the README file to include comments on t/20-certificate.t and what
    changes are required before you can use it.
  - Updated t/20-certificate.t to highlight the section of code that the README
    file is talking about and the initial prompt to remind the user to see the
    the README file for the needed modifications before running the test!

0.18 2011/07/29 08:30:00
  - Added SSL_Client_Certificate to support client certificates on the data
    channel as well as the command channel. Bug ID: 69327.
  - Depreciated SSL_Advanced in favor of SSL_Client_Certificate.  SSL_Advanced
    is now just an alias for SSL_Client_Certificate for backwards compatibility.
    I just feel the new name better describes what we are trying to do with it
    now that this feature is supported.
  - More fixes/clarifications to the POD text.
  - Fixed minor wild card issue with LIST & NLST if "+" was in the file name.
  - Added restart() function to be compatable with Net::FTP for restarts.
  - Added ability to restart downloads/get() where previous attempt left off.
  - Added ability to restart uploads/put() where the previous attempt left off.
  - Added append() command [APPE], it also uses OFFSET, but doesn't send the
    REST command to the server.
  - Added LocalAddr option to new().  Works against both the command channel
    and the data channel.
  - Masks out the user name used when generating a log file.  Similar to how
    this is already done for the password.  Except it must also hide it in the
    response to USER & PASS.  Only masked when written to the log, message()
    and last_message() still returns the user name unmasked in any response
  - Fixed a bug in t/10-complex.t and added optional environment variables
    for some of the connection values to make it easier for me to quickly
    rerun tests during development without having to key in ansers to questions
    with no good default values over & over again.
  - Wrote t/20-certificate.t to test certificates out.  But this and
    t/10-complex.t are probably incompatable in most settings.  Anyone else
    trying to use this test case will have to modify this code to point to their
    certificate and provide it's password, since I'm not going to publish my
    test client certificate.
  - Calls to command() now return a reference to this class so that you can
    code something like: "if ($ftps->command("NOOP")->response() == CMD_OK)".
    This should be compatible since it shouldn't be called externally, and
    you should always be calling response() afterwards anyway.  It still calls
    die if "Croak=>1" is used.  Response() will return CMD_ERROR should
    command() have issues to avoid hangs.
  - Modifed Net::FTPSSL to use the new command()->response() format internally.

0.17 2011/01/03 08:30:00
  - Fixed a subtle bug in response(), now tests if a parse succeeded instead
    of assuming it.  Allowed me to simplify this messy code quite a bit.
  - Reordered the methods in the POD to put some of the more relevant FTP
    commands closer to the top & the more obscure at the bottom to make
    for easier reading of the documentation.
  - Reworked t/10-complex.t quite a bit.  Made sure croak was always turned
    off immediately after a successful login.  Also now using the same hash
    for all the connection attempts.  I really need to take the time to do
    further redesign this script to make it easier to troubleshoot.
  - Fixed the smoke tester failures.  (They always answer NO to the deeper test)
  - t/10-complex.t now checks if OverridePASV or OverrideHELP are needed against
    the test server & dynamically adds it for the test cases if needed.
    Net::FTPSSL still can't dynamically figure this out for itself.
  - Added PASV/EPSV prompt to t/10-complex.t script instead of assuming PASV.

0.16 2010/11/30 08:30:00
  - Used Perl's special variables to write the Perl verion and OS to the logs
    to make support easier when trouble shooting tickets.
  - Fixes for Bug Id 61432.
  - On systems like "Connect:Enterprise UNIX", it incorrectly sends the
    response to HELP partially in clear text and partially encrypted on the
    command channel.  This causes this class to barf and never recover once it
    hits the clear text.  So a new "OverrideHELP" option was created to bypass
    calls to HELP if your server has this strange behavior.  This is needed
    since HELP is called internally by the Net::FTPSSL class via supported()!
    ** Maybe in the future I'll figure out a way to dynamically handle this. **
  - On some systems the server administrator won't configure their servers
    correctly to use the correct IP Address in the PASV response through a
    firewall when establishing a new Data Channel Connection.  To get arround
    this, a new "OverridePASV" option was created to force the use of the
    Specified Server Name instead of the invalid IP Address returned in the
    PASV response!
    ** Maybe in the future I'll figure out a way to dynamically handle this. **
  - Added "OverridePASV" & "OverrideHELP" test cases to t/10-complex.t
  - Added EPSV support.  Required the reorginization of some internal
    undocumented methods that no one should be using.
  - Added EPSV test to t/10-complex.t

0.15 2010/04/27 08:30:00
  - Fixes for Bug Id 53719.  Requested all internal calls to "ALLO" be
    conditional since some server connections die when it receives an "ALLO"
  - Fixing the above bug uncovered an infinite loop if Croak was also turned on.
    Now _croak_or_return() doesn't try to close the connection if called by
    command() or response() or if recursion was detected.
  - Fixed quot() to no longer upshift the command to upper case.  Discovered
    case where lower case was required.  So it's now up the the caller to
    put in upper case if it's required!  Also no longer checks if the command
    is valid according to HELP.  Some servers attempt to be sneeky by not
    advertising all available commands.

0.14 2010/01/04 08:30:00
  - Added minimum requirement of IO::Socket::SSL v1.08, since I just got word
    that v0.97 caused Net-FTPSSL to hang!  And v1.08 was the release I
    originally developed Net::FTPSSL with & is easily available for Windows
    users as a pre-built release. But that release is still pretty old.
    (IO::Socket::SSL is currently at v1.31 as of this writting.)
  - Fixed bug in _common_put() where it didn't actually test the final status
    message when closing the data connection.  So missed a rare failure case.
  - Fixed similar bugs in list() & get() as well.
  - Caught more list() & nlst() error conditions that returned undef instead of
    the empty list as documented.
  - Added the Server/Port to start of trace file after the version # since not
    all servers returned this info in their log messagse.

0.13 2009/11/01 08:30:00
  - Decided to skip unlucky release # 13!

0.12 2009/09/29 08:30:00
  - Made some POD changes.
  - Fixed so each class instance has it's own file hande so multiple instances
    won't write to the log file of the last Net::FTPSSL->new() call made with
    Debug turned on.  Also fixed t/10-complex.t to use multiple connections so
    we can manually verify this is finally corrected.
  - Made sure the version of FTPSSL used always gets written out when Debug is
    turned on.  So removed this code from t/10-complex.t as no longer needed.
  - Fixed "quit" to close the DebugLogFile, will write to STDERR if anything
    else is improperly called after "quit" is called!
  - Fixed "put" tie file handle problem from Tkt # 49378.
  - Added mdtm() & mfmt() to follow FTP specs & exposed _mdtm() & _mfmt() in
    the POD text, which uses timestamps instead.  Per request # 49376.  The
    underscore versions are more user friendly.
  - Added size() per request # 49377.
  - Minor fixes to 10-complex.t to support new functionality.
  - Reorged some older code to make it easier to maintain.  Making some blocks
    of code much smaller and less awkward.
  - Fixed xput & xget to delete the dest file before renaming the scratch file.
    Avoids rare file permission issues when the dest file already exists and the
    file recognizer is down.

0.11 2009/07/19 08:30:00
  - Added ccc() to end of 10-complex.t test script.
  - Updated the ccc() docs on issues with older IO::Socket::SSL versions.
  - Provided solution to Tkt # 47659, problems with CCC - RFC 4217 Section 12.3.
    Where CCC behaves differently for different FTPS servers.
  - Added Debug option to allow writing to DebugLogFile in append mode instead
    of creating a new log file every time.  So can use same log file when
    serially making FTPS connections.
  - Now allows overriding of the IMP_CRYPT port 990 to something else.  Found a
    server that allowed me to change the implicit port # for this. Tkt # 46851.

0.10 2009/06/30 08:30:00
  - Updated the Port vs IMP_CRYPT documentation to state you can't override
    port 990 for IMP_CRYPT.  Where before this was just assumed.
  - Added option DebugLogFile to allow you to write your Debug info to a
    file of your choice instead of STDERR.  This way multiple FTPS connections
    can be traced separately of each other.  Also keeps other writes to STDERR

0.09 2009/05/13 08:30:00
  - Fixed some typos in the documentation.
  - Fixed so list() & nlst() no longer return undef on some errors instead of
    the empty list as documented.
  - Fixed bug # 45710 where nlst() has issues returning zero rows.
  - Fixed xput() so the scratch file is guaranteed to appear in the same
    directory as the final file unless the prefix overrides it with a directory
    of it's own.  Fixes issue where the remote file wasn't being written to the
    current directory on the remote server and we couldn't write to the current
    directory on that server for the scratch file.
  - Added xget() as the complement to xput().  Where the file recognizer is on
    the client side.
  - Added ccc() for finally supporting the CCC command. (Clear Command Channel)
  - Improved get() logic for removing zero byte files if it can't download the
    requested file.  Still leaves behind partially downloaded files on purpose.
  - Enhanced t/10-complex.t to test new functionality. (But not CCC due to
    security concerns.)
  - Added last_status_code() to return the 1 digit status returned by message()
    so your code can branch on particular issues instead of just pass/fail.

0.08 2009/03/23 08:30:00
  - Fixed new() to also accept a hash reference.  So can do either way now!
    Pass the hash by reference or pass the hash by value.
  - Fixed command() & response() undefined function bug when called by a
    socket data type in new() and a fatal error was encountered.  Solved by
    not calling the member function in the normal way.
  - Added $Net::FTPSSL::ERRSTR to give you access to the error messages
    generated when you couldn't create a Net::FTPSSL object via new().  Also
    set when Croak is called even though you had the special perl variable $@
    for this. It is also printed to STDERR when Debug is turned on.
  - Fixed response() case where the ending message had CR's in it and it was
    being truncated.
  - Fixed response() to detect unexpected EOF: Bug # 43670.
  - Now passes Timeout to start_SSL() calls in new().
  - Added plain old FTP as option CLR_CRYPT.  Just avoids encrypting the
    command channel.  Still doesn't support the CCC command.
  - Fixed implicit problem.  Turned out to be a bug in choosing the "default"
    port as well as reading the response to soon. Bug # 28914.
  - Exposed all the FTP CMD status constants for public use.
  - Added unsupported option SSL_Advanced for Enhancement Request # 44042.
    Use at your own risk!  It's not supported by the developer of Net::FTPSSL.
  - Enhanced t/10-complex.t to print the version of Net-FTPSSL being tested and
    support/test the new functionality.  Also now generates a backup copy of
    the trace log named after the options selected to ease testing of multiple
  - Added xput for avoiding file recognizer issues on the FTPS server side
    during file transfers.  IE the file recognizer picks the file up before the
    transfer completes.  After the rename the file recognizer can safely assume
    the file transfer has completed without issues.
  - Added option in new() to support preserving the timestamps on files
    transfered between the client & server via get(), put(), uput() & xput().
    Works providing the FTPS server supports this functionality.
  - Fixed response() timing bug where sysread() sometimes read the results of
    multiple commands.  Ex: the 150 INFO msg for opening the data connection and
    the 226 transfer complete message.  This bug caused FTPSSL to randomly hang
    when connected to some servers.
  - Fixed bug where nlst() hung if it returned zero rows on some servers.
  - Removed total from list() since it wasn't always present for all servers
    and it was the total block size, not the total # of files or bytes returned.
  - Fixed list() & nlst() to allow wildcard filters similar to unix "ls" cmd.
    The only wildcards being "*" or "?".

0.07 2009/02/24 08:30:00
  - Corrected some typos in the documentation.
  - Implemented call back functionality for all data channel functions.
  - Fixed uput() to return the file name used on success instead of just
    true/false.   Will still return "undef" on failure.  Needed in order to
    figure out what the other server actually called the file we sent it.
  - Exposed the DataProtLevel constants for public use.
  - Added callback feature for all data channel FTP functions.
  - Now uses caller() extensively to combine multiple similar functions together
    and for callback support.
  - Another fix to the get() bug introduced by the Bug 17537 fix.  The previous
    attempt in v0.06 didn't fully fix the problem!  So took different approach.
  - Enhanced t/10-complex.t & t/00-basic.t

  The following changes may break some existing code ...
  - Added new option 'Croak' to cause all failures to call croak() instead of
    returning undef.  This caused some minor incompatabilities on error handling
    if your code depended on some old croak() calls.  But the new code is now
    consistant in error handling in either always returning failure or always
    calling croak!  It no longer does a combination of both.
  - Added set_croak() to help mitigate above issue, by being able to turn the
    croak feature on & off.  t/10-complex.t uses this logic while initializing
    the connection.
  - user() & password() have been renamed since internal functions.

0.06 2009/02/03 08:30:00
  - Added new() DataProtLevel option to allow selecting the Data Channel
  - Fixed _help() so supported() works for some new servers.
  - new() now allows you to select SSL over TLS for connections via useSSL.
  - Fixed get() bug that sometimes added extra \015 to ASCII files downloaded.
    Introduced when Bug 17537 was fixed.
  - Enhanced t/10-complex.t

0.05 2009/01/05 08:30:00
  - Fixed resonse() to properly get the entire response instead of just the 1st
    line of it.  This change fixes many of the reported bugs reported against
    this module.
  - Fixed command() & response() to also log socket() calls in debug mode with
    "SKT >>>" & "SKT <<<" prefixes.
  - Added supported(), quot(), & _help()
  - All response calls in new() are now tested & added debug flag to socket
  - Enhanced t/10-complex.t
  - Other minor fixes.
  - Bugs Fixed: 41665, 31720, 16751, 30359, 24136, 17537, 17538, 34818

Version - DD/MM/YYYY -- The Marco Dalla Stella releases.

0.04 23/11/2005 11:58:34
  - Corrected some typos on the documentation.

0.03 05/09/2005 15.01.52 
  - cdup(), mkdir() and rmdir() added
  - A better sysread() and syswrite() use on list(), nlst(), get(), put() and
    uput() methods
  - Added autoflush on list(), nlst(), get(), put() and uput() methods
  - A better interactive test, (the previous one sucked, this one is a little
    better! ^_^)
  - Other minor fixes

0.02 01/08/2005 22:12:34
  - Change the returning data of list() and nlst()
  - Fix ALLO command call in put() and uput() methods
  - Correct the PASV parsing regexp (thanks to sfr)
  - Makefile no longer require Perl 5.8.4

0.01 21/05/2005 20:55:58
  - original version; created by h2xs 1.23 with options -AX Net::FTPSSL