CHANGES

  The problem is the following: when we're calling Curl_done and it decides to
  keep the connection opened ('left intact'), then the caller is not notified
  that the connection was done via the NULLifying of the pointer, so some easy
  handle is keeping the pointer to this connection.

  Later ConnectionExists can select such connection for reuse even if we're
  not pipelining: pipeLen is zero, so the (pipeLen > 0 && !canPipeline) is
  false and we can reuse this connection for another easy handle. But thus the
  connection will be shared between two easy handles if the handle that wants
  to take the ownership is not the same as was not notified of the connection
  was done in Curl_done. And when some of these easy handles will get their
  connection really freed the another one will still keep the pointer.

  My fix was rather trivial: I just added the NULLification to the 'else'
  branch in the Curl_done. My tests with Git and ElectricFence showed no
  problems both for HTTP pulling and cloning. Repository size is about 250 Mb,
  so it was a considerable amount of Curl's work.

Dan F (9 March 2007)
- Updated the test harness to add a new "crypto" feature check and updated the
  appropriate test case to use it.  For now, this is treated the same as the
  "SSL" feature because curl doesn't list it separately.

Daniel (9 March 2007)
- Robert Iakobashvili fixed CURLOPT_INTERFACE for IPv6.

- Robert A. Monat improved the maketgz and VC6/8 generating to set the correct
  machine type too.

- Justin Fletcher fixed a file descriptor leak in the curl tool when trying to
  upload a file it couldn't open. Bug #1676581
  (http://curl.haxx.se/bug/view.cgi?id=1676581)

Dan F (9 March 2007)
- Updated the test harness to check for protocol support before running each
  test, fixing KNOWN_BUGS #11.

Dan F (7 March 2007)
- Reintroduced (after a 3 year hiatus) an FTPS test case (400) into the test
  harness.  It is very limited as it supports only ftps:// URLs with
  --ftp-ssl-control specified, which implicitly encrypts the control
  channel but not the data channels.  That allows stunnel to be used with
  an unmodified ftp server in exactly the same way that the test https
  server is set up.

Dan F (7 March 2007)
- Honour --ftp-ssl-control on ftps:// URLs to allow encrypted control and
  unencrypted data connections.

Dan F (6 March 2007)
- Fixed a couple of improper pointer uses detected by valgrind in test
  cases 181 & 216.

Daniel (2 March 2007)
- Robert A. Monat and Shmulik Regev helped out to fix the new */Makefile.vc8
  makefiles that are included in the source release archives, generated from
  the Makefile.vc6 files by the maketgz script. I also modified the root
  Makefile to have a VC variable that defaults to vc6 but can be overridden to
  allow it to be used for vc8 as well. Like this:

    nmake VC=vc8 vc

Daniel (27 February 2007)
- Hang Kin Lau found and fixed: When I use libcurl to connect to an https
  server through a proxy and have the remote https server port set using the
  CURLOPT_PORT option, protocol gets reset to http from https after the first
  request.
 
  User defined URL was modified internally by libcurl and subsequent reuse of
  the easy handle may lead to connection using a different protocol (if not
  originally http).
 
  I found that libcurl hardcoded the protocol to "http" when it tries to
  regenerate the URL if CURLOPT_PORT is set. I tried to fix the problem as
  follows and it's working fine so far

Daniel (25 February 2007)
- Adam D. Moss made the HTTP CONNECT procedure less blocking when used from
  the multi interface. Note that it still does a part of the connection in a
  blocking manner.

Daniel (23 February 2007)
- Added warning outputs if the command line uses more than one of the options
  -v, --trace and --trace-ascii, since it could really confuse the user.
  Clarified this fact in the man page.

Daniel (21 February 2007)
- Ravi Pratap provided work on libcurl making pipelining more robust and
  fixing some bugs:
  o Don't mix GET and POST requests in a pipeline
  o Fix the order in which requests are dispatched from the pipeline
  o Fixed several curl bugs with pipelining when the server is returning
    chunked encoding:
    * Added states to chunked parsing for final CRLF
    * Rewind buffer after parsing chunk with data remaining
    * Moved chunked header initializing to a spot just before receiving
      headers

Daniel (20 February 2007)
- Linus Nielsen Feltzing changed the CURLOPT_FTP_SSL_CCC option to handle
  active and passive CCC shutdown and added the --ftp-ssl-ccc-mode command
  line option.

Daniel (19 February 2007)
- Ian Turner fixed the libcurl.m4 macro's support for --with-libcurl.

- Shmulik Regev found a memory leak in re-used HTTPS connections, at least
  when the multi interface was used.

- Robson Braga Araujo made passive FTP transfers work with SOCKS (both 4 and
  5).

Daniel (18 February 2007)
- Jeff Pohlmeyer identified two problems: first a rather obscure problem with
  the multi interface and connection re-use that could make a
  curl_multi_remove_handle() ruin a pointer in another handle.

  The second problem was less of an actual problem but more of minor quirk:
  the re-using of connections wasn't properly checking if the connection was
  marked for closure.

Daniel (16 February 2007)
- Duncan Mac-Vicar Prett and Michal Marek reported problems with resetting
  CURLOPT_RANGE back to no range on an easy handle when using FTP.

Dan F (14 February 2007)
- Fixed curl-config --libs so it doesn't list unnecessary libraries (and
  therefore introduce unnecessary dependencies) when it's not needed.
  Also, don't bother adding a library path of /usr/lib

Daniel (13 February 2007)
- The default password for anonymous FTP connections is now changed to be
  "ftp@example.com".

- Robert A. Monat made libcurl build fine with VC2005 - it doesn't have
  gmtime_r() like the older VC versions. He also made use of some machine-
  specific defines to differentiate the "OS" define.

Daniel (12 February 2007)
- Rob Crittenden added support for NSS (Network Security Service) for the
  SSL/TLS layer. http://www.mozilla.org/projects/security/pki/nss/

  This is the fourth supported library for TLS/SSL that libcurl supports!

- Shmulik Regev fixed so that the final CRLF of HTTP response headers are sent
  to the debug callback.

- Shmulik Regev added CURLOPT_HTTP_CONTENT_DECODING and
  CURLOPT_HTTP_TRANSFER_DECODING that if set to zero will disable libcurl's
  internal decoding of content or transfer encoded content. This may be
  preferable in cases where you use libcurl for proxy purposes or similar. The
  command line tool got a --raw option to disable both at once.

- release tarballs made with maketgz will from now on have a LIBCURL_TIMESTAMP
  define set to hold the exact date and time of when the tarball was built, as
  a human readable string using the UTC time zone.
  
- Jeff Pohlmeyer fixed a flaw in curl_multi_add_handle() when adding a handle
  that has an easy handle present in the "closure" list pending closure.

Daniel (6 February 2007)
- Regular file downloads wiht SFTP and SCP are now done using the non-blocking
  API of libssh2, if the libssh2 headers seem to support them. This will make
  SCP and SFTP much more responsive and better libcurl citizens when used with
  the multi interface etc.

Daniel (5 February 2007)
- Michael Wallner added support for CURLOPT_TIMEOUT_MS and
  CURLOPT_CONNECTTIMEOUT_MS that, as their names suggest, do the timeouts with
  millisecond resolution. The only restriction to that is the alarm()
  (sometimes) used to abort name resolves as that uses full seconds. I fixed
  the FTP response timeout part of the patch.

  Internally we now count and keep the timeouts in milliseconds but it also
  means we multiply set timeouts with 1000. The effect of this is that no
  timeout can be set to more than 2^31 milliseconds (on 32 bit systems), which
  equals 24.86 days.  We probably couldn't before either since the code did
  *1000 on the timeout values on several places already.

Daniel (3 February 2007)
- Yang Tse fixed the cookie expiry date in several test cases that started to
  fail since they used "1 feb 2007"...

- Manfred Schwarb reported that socks5 support was broken and help us pinpoint
  the problem. The code now tries harder to use httproxy and proxy where
  apppropriate, as not all proxies are HTTP...

Version 7.16.1 (29 January 2007)

Daniel (29 January 2007)
- Michael Wallner reported that when doing a CONNECT with a custom User-Agent
  header, you got _two_ User-Agent headers in the CONNECT request...! Added
  test case 287 to verify the fix.

Daniel (28 January 2007)
- curl_easy_reset() now resets the CA bundle path correctly.

- David McCreedy fixed the Curl command line tool for HTTP on non-ASCII
  platforms.

Daniel (25 January 2007)
- Added the --libcurl [file] option to curl. Append this option to any
  ordinary curl command line, and you will get a libcurl-using source code
  written to the file that does the equivalent operation of what your command
  line operation does!

Dan F (24 January 2007)
- Fixed a dangling pointer problem that prevented the http_proxy environment
  variable from being properly used in many cases (and caused test case 63
  to fail).

Daniel (23 January 2007)
- David McCreedy did NTLM changes mainly for non-ASCII platforms:

  #1
  There's a compilation error in http_ntlm.c if USE_NTLM2SESSION is NOT
  defined.  I noticed this while testing various configurations.  Line 867 of
  the current http_ntlm.c is a closing bracket for an if/else pair that only
  gets compiled in if USE_NTLM2SESSION is defined.  But this closing bracket
  wasn't in an #ifdef so the code fails to compile unless USE_NTLM2SESSION was
  defined.  Lines 198 and 140 of my patch wraps that closing bracket in an
  #ifdef USE_NTLM2SESSION.

  #2
  I noticed several picky compiler warnings when DEBUG_ME is defined.  I've
  fixed them with casting.  By the way, DEBUG_ME was a huge help in
  understanding this code.

  #3
  Hopefully the last non-ASCII conversion patch for libcurl in a while.  I
  changed the "NTLMSSP" literal to hex since this signature must always be in
  ASCII.

  Conversion code was strategically added where necessary.  And the
  Curl_base64_encode calls were changed so the binary "blobs" http_ntlm.c
  creates are NOT translated on non-ASCII platforms.

Dan F (22 January 2007)
- Converted (most of) the test data files into genuine XML.  A handful still
  are not, due mainly to the lack of support for XML character entities
  (e.g. & => & ).  This will make it easier to validate test files using
  tools like xmllint, as well as to edit and view them using XML tools.

Daniel (16 January 2007)
- Armel Asselin improved libcurl to behave a lot better when an easy handle
  doing an FTP transfer is removed from a multi handle before completion. The
  fix also fixed the "alive counter" to be correct on "premature removal" for
  all protocols.

Dan F (16 January 2007)
- Fixed a small memory leak in tftp uploads discovered by curl's memory leak
  detector.  Also changed tftp downloads to URL-unescape the downloaded
  file name.

Daniel (14 January 2007)
- David McCreedy provided libcurl changes for doing HTTP communication on
  non-ASCII platforms. It does add some complexity, most notably with more
  #ifdefs, but I want to see this supported added and I can't see how we can
  add it without the extra stuff added.

- Setting CURLOPT_COOKIELIST to "ALL" when no cookies at all was present,
  libcurl would crash when trying to read a NULL pointer.

Daniel (12 January 2007)
- Toby Peterson found a nasty bug that prevented (lib)curl from properly
  downloading (most) things that were larger than 4GB on 32 bit systems.  Matt
  Witherspoon helped as narrow down the problem.

Daniel (5 January 2007)
- Linus Nielsen Feltzing introduced the --ftp-ssl-ccc command line option to
  curl that uses the new CURLOPT_FTP_SSL_CCC option in libcurl. If enabled, it
  will make libcurl shutdown SSL/TLS after the authentication is done on a
  FTP-SSL operation.

Daniel (4 January 2007)
- David McCreedy made changes to allow base64 encoding/decoding to work on
  non-ASCII platforms.

Daniel (3 January 2007)
- Matt Witherspoon fixed the flaw which made libcurl 7.16.0 always store
  downloaded data in two buffers, just to be able to deal with a special HTTP
  pipelining case. That is now only activated for pipelined transfers. In
  Matt's case, it showed as a considerable performance difference,

Daniel (2 January 2007)
- Victor Snezhko helped us fix bug report #1603712
  (http://curl.haxx.se/bug/view.cgi?id=1603712) (known bug #36) --limit-rate
  (CURLOPT_MAX_SEND_SPEED_LARGE and CURLOPT_MAX_RECV_SPEED_LARGE) are broken
  on Windows (since 7.16.0, but that's when they were introduced as previous
  to that the limiting logic was made in the application only and not in the
  library). It was actually also broken on select()-based systems (as apposed
  to poll()) but we haven't had any such reports. We now use select(), Sleep()
  or delay() properly to sleep a while without waiting for anything input or
  output when the rate limiting is activated with the easy interface.

- Modified libcurl.pc.in to use Libs.private for the libs libcurl itself needs
  to get built static. It has been mentioned before and was again brought to
  our attention by Nathanael Nerode who filed debian bug report #405226
  (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=405226).