CHANGELOG

Version 2.9 12/07/2019
======================

 * Updates to project dependencies, to handle upstream Selenium project changes (contributed by https://github.com/DeathBeforeDecaf)
 * Bundle Apache HTTP client and commons-logging, since Selenium no longer depends on these projects.

Version 2.8 14/08/2018
======================
 * Update to fix compile errors when running against newer versions of Guava (related to the Files.fileTreeTraverser() API)

Version 2.7 08/04/2018
=======================
 * Add JMX annotations, so the VideoProxy still works under Selenium 3.9.0

Version 2.6 19/11/2017
======================

 * Update project to build against upstream code changes - Selenium 3.7.1 now required (Use version 2.5 for Selenium 3.6.0 and below).
 * Expand unit test coverage for hub proxy

Version 2.5 11/11/2017
======================

 * Bump build to Java 8 bytecode
 * Refactor to make unit testing easier
 * Add some unit tests
 * Fix rounding bug in FPS calculations
 * Bundle JNA dependencies in JAR file - recent versions of Selenium no longer use this

Version 2.4 01/10/2017
======================

 * Update version dependencies a bit (contributed by saikek)
 * Add AWS video upload support (contributed by ovidiubute)

Version 2.3 26/02/2017
======================

 * Fix failure introduced by Selenium 3.1 in the capabilities setting code (issue #27, issue #35)
 * Greatly improve X11 screenshot performance by writing a greatly simplified version of the AWT Robot code (issue #14)
   This is off by default for now, but can be enabled in Linux/Unix by passing "-Dvideo.source=x11" to the JVM

Version 2.2 01/01/2017
=======================

 * Re-add support for Selenium V2, by using Java reflection to smooth over incompatible changes (issue #27)

Version 2.1 01/11/2016
=======================

 * Properly fix issue #23, where JNA was picking up the system version of libvpx on Linux, causing the JVM to segfault.

Version 2.0 01/11/2016
=======================

 * Update build to support Selenium 3 (contributed by mhardin)

Version 1.9 23/12/2015
======================

 * Recompile 32-bit Windows native dependencies, to fix problem where they were dynamically linked against GCC's libc library (issue #19)
 * Commit fix for issue #5 (where the node can get into an inconsistent recording state if the hub gets restarted)
 * Logging/documentation improvements

Version 1.8 26/06/2015
======================

 * A lot of POM changes to upload this project to Maven central (no code changes)
 * Fix issue where the native code zip file was being included in the source JAR file

Version 1.7 12/02/2014
======================

* Fix for issue where dynamically changing resolution before recording a test didn't record at the new resolution - we used the 
  resolution at startup (fix contributed by https://github.com/suayipozmen)

Version 1.6 05/02/2014
======================

* Fix issue #10 - make video download requests wait until the Hub has retrieved the video from the Node


Version 1.5 22/11/2014
======================

* Bundled the org.json:json Maven artifact, so we can parse JSON from Selenium 2.44 onwards - they dropped it as a dependency (issue #9)
* Fix integer promotion bug in the xvfb acceleration code, which allows us to record in larger resolutions (fix contributed by https://github.com/DeathBeforeDecaf)
* Bump Selenium version we are compiling against to 2.43 (shouldn't make any difference)

Version 1.4 21/11/2014
======================

* Add pre-compiled and tested binaries for 32-bit Windows
* Add check in Java code to make sure we are taking screenshots with the correct bit depth

Version 1.3 11/09/2014
======================

* Make the node delete extracted native dependencies and recorded videos on exit (fixes issue #7).
  This could cause the machine the node was running on to diskfill if the node was repeatedly restarted.

Version 1.2 15/08/2014
======================

* Add a new means of taking screenshots for Xvfb, where we can memory-map the Xvfb output into our address space to run the video encoder directly on.
  Note that this requires an extra JVM flag to make this work.
* Check that the video data source is sane on startup, instead of when we first record a video.

Version 1.1 8/07/2014
=====================

* Add an API call to the Node to reset the video recording to a known state (will be used in future releases)
* Fix some resource leaks in Video Download/Info servlets (issue #4)

Version 1.0 1/07/2014
=====================

* Bump to release version (no code changes)

Version 0.11 29/06/2014
=======================

* Handle native code error conditions better (but haven't managed to test it because it works perfectly :/)
* Tweak the IVideoStore interface again to also pass in the video content length (in bytes)
* Reduce the excess video recording time when closing the session to be 1/2 a second (was 1 second)
* Plug a security hole which existed in most of the video storage mechanisms by which a request could access any file ending in .webm off the filesystem.

Version 0.10 25/06/2014
=======================

* Fix an issue where HTTP connections leaked, and we couldn't acquire any more
* Change the native code extraction logic to only happen once at startup (previously we also did it when recording a video too)
* Bump the default framerate to 15 - some platforms can record faster than others.

Version 0.9 24/06/2014
======================

* Recompiled Windows x86_64 binaries to fix video length bug (Issue #3)
* Changed the interface for IVideoStore to be a bit cleaner

Version 0.8 23/06/2014
======================

* Added pre-compiled binaries for OS X
* Added pre-compiled binaries for Linux x86 (32-bit)
* Recompiled all binaries (except Windows x86_64) to fix video length bug (Issue #3)
* Change the API for IVideoStore to also include requested capabilities + node capabilities when storing the job. This is intended
for use when integrating with custom job reporting code.
* Renames packages/POM from 'com.mooo.aimmac23' -> 'com.aimmac23'
* Removed experimental support for Windows x86 (32-bit). 64-bit binaries compiled by the same process crashed, so these probably won't work either.

Version 0.7 20/06/2014
======================

* Added experimental support for Windows x86 (32-bit). I don't have a 32-bit JVM on-hand to test that the native code works.
* Wrote a plugin to make how we store videos be extensible
  * LocalTempFileStore - the pre-existing storage mechanism
  * LocalFileVideoStore - permanent storage in a directory
  * BasicWebDAVStore - permanent storage in a WebDAV directory on a webserver
* Added HubVideoInfoServlet to provide storage-plugin specific information about where a video is kept

Version 0.6 26/05/2014
======================

* Fixed an issue where videos were not immediately available after the session was closed - only after the Hub had finished tidying up the node. We now immediately transfer the video to the Hub.

Version 0.5 03/05/2014
======================

* Make the target framerate be configurable
* Fix a playback framerate issue when the video wasn't recorded fast enough (was playing too fast)
* Recompile binaries on glibc 2.13, because not all Linux installations have 2.17

Version 0.4 21/04/2014
======================

* Commit some pre-built linux binaries to Git.
* Make Maven assemble the native code zip file.
* Added support for Windows x86_64 - committed some binaries for it (details on wiki). Note that this requires a 64 bit JVM to work.

Version 0.3 18/04/2014
======================

* Move the compiled MKV encoder code out into its own library, where it should be.
* Fix C memory leak by actually freeing up memory.


Version 0.2 15/04/2014
======================

Convert project to use a WebM/VP8 encoder - we now use JNA to run the encoder, with a C implementation to abstract away the complicated C data structures.

 * Higher framerates achieved
 * Capped framerate at 8 fps, slightly slower than the maximum speed, due to continuing time-base issues
 * Implementation now restricted to Linux-only.


Version 0.1 08/04/2014n
======================

Initial implementation. Used a Java implementation of an h264 encoder. It centralized videos on the Hub using some download servlets, and allowed the user to get the video.

 * The encoder wasn't very fast (~4 fps, with the Java Robot taking screenshots)
 * The videos generated were massive
 * Some time-base issues
 * Some browsers/players refused to play back the video
 * Patent issues for encoding h264