Updated the performance chapter.

Own Id: OTP-8696

A minor compiler related performance improvement.

Own Id: OTP-8561

A race condition when, during high load, processing both the original and a resent message and delivering this as two separate messages to the user.

Note that this solution only protects against multiple reply deliveries!

Own Id: OTP-8529

Aux Id: Seq 10915

Fix shared libraries installation.

The flex shared lib(s) were incorrectly installed as data files.

Peter Lemenkov

Own Id: OTP-8627

Eliminated a possible race condition while creating pending counters.

Own Id: OTP-8634

Aux Id: Seq 11579

Various changes to configure and makefile(s) to facilitate cross compilation (and other build system improvements).

Own Id: OTP-8323

Added a help target in the test Makefile to explain the most useful make targets, used when testing the application using the test-server provided with megaco.

Own Id: OTP-8362

Adapted megaco_filter to the new internal format.

Own Id: OTP-8403

Callbacks, when the callback module is unknown (undefined), results in warning messages.

A race condition scenario. As part of a cancelation operation, replies with waiting acknowledgements is cancelled. This includes informing the user (via a call to the handle_trans_ack callback function). It is possible that at this point the connection data has been removed, which makes it impossible for megaco to perform this operation, resulting in the warning message. The solution is to also store the callback module with the other reply information, to be used when cleaning up after a cancelation.

Own Id: OTP-8328

Aux Id: Seq 11384

A minor optimization by using ets:update_element instead of ets:insert for some table updates.

Own Id: OTP-8239

The documentation is now built with open source tools (xsltproc and fop) that exists on most platforms. One visible change is that the frames are removed.

Own Id: OTP-8249

Improve handling of async transaction reply.

For asynchronous requests, issued using megaco:cast/3, the reply will be delivered using the handle_trans_reply/4,5 callback function.

If a receiver of a request, issued using megaco:cast/3, does not reply in time, megaco re-sends the request. If the receiver of the request sends the reply at the same time as megaco re-sends, it may also send a reply to the resent request (thinking the first reply got lost). These two replies may arrive more or less at the same time, causing confusion.

In order to improve this situation, a number of improvements have been done:

Own Id: OTP-8183

Aux Id: Seq 11393

If the megaco app received a transaction reply, for a request issued using the call/3 function, from the wrong remote entity (wrong MId)), megaco would still deliver the reply (call/3 returnes) as if from the correct remote entity (right MId).

This has been changed so that the function now returns with an error reason.

See call/3 for more info.

*** POTENTIAL INCOMPATIBILITY ***

Own Id: OTP-8212

Aux Id: Seq 11305

Replacing obsolete guard tests.

Own Id: OTP-8164

Added the config option call_proxy_gc_timeout to be able to control the way unexpected replies (when requests issued via calls to call/3) are handled.

See user_info/2, conn_info/2 and call/3 for more info.

Own Id: OTP-8167

Aux Id: Seq 11393

Make flex scanner c89 compiler compliant.

Akira Kitada

Own Id: OTP-8191

Megaco was unnecessarily strict when parsing the SDP attribute maxptime (leading or trailing spaces cased the value parse to fail).

This has been improved so that leading and trailing spaces are stripped before parsing the value. The same has been done for the attribute ptime.

Own Id: OTP-8123

Aux Id: Seq 11364

[text] The unquoted string BOTH was interpreted as the 'BothToken' token. This was a version 3 (prev3a, prev3b, prev3c and v3) only.

Own Id: OTP-8114

Aux Id: Seq 11353

The reply proxy could crash if the timeout time calculation results in a negative number. This will result in a function clause with resulting error report.

Own Id: OTP-8081

Aux Id: Seq 11332

In order to better utilize multi-core procs, the flex (text) scanner has been improved.

The flex (text) scanner has been made reentrant, if the flex utility supports this. Note that the version of flex supplied with some OS/distros (Solaris 10, FreeBSD and OpenBSD to mention a few) may not support this, in which case the flex scanner will be non-reentrant, just as before.

Own Id: OTP-7302

Unexpected handle_unexpected_reply callbacks.

The megaco_user callback function handle_unexpected_reply could during high load be called with unexpected values for the Trans argument, such as an TransactionReply where transactionResult had the value {error, timeout}. This was a result of a race condition and has now been fixed.

Own Id: OTP-7926

Aux Id: Seq 11255

[text] PropertyParm values cannot be quoted.

It was not possible to encode a PropertyParm value as a quoted string (unless it *had* to (has at least one RestChar)). The megaco text codec's now also accepts quoted strings as PropertyParm values.

Own Id: OTP-7936

Aux Id: Seq 11258

Updated file headers.

Own Id: OTP-7851

Added new API function megaco:connect/5 and the corresponding new megaco_user callback function handle_connect/3. The purpose of this is to be able to pass information to the handle_connect/3 function by calling the megaco:connect/5 function.

Own Id: OTP-7713

Aux Id: Seq 11140

Update file headers with new copyright notice.

Own Id: OTP-7743

If the transport module calls the process_received_message/5 or receive_message/5 function(s) for the initial message, then the handle_connect/3 function will now be called and not the handle_connect/2 function.

Own Id: OTP-7713

Aux Id: Seq 11140