native-processes-not-closing -- Native processes should be closed when not used

[deyan]

Media decoders and browser embed should be closed after a timeout when not used. Browser embed should be closed after the book containing brwoser frame is closed.

[meddle]

The analysis is ok for me 4p

[meddle]

The design:

• All the NativeBridges in sophie will be registered in a special NativeBridgeRegistry.
  • The registry will have a daemon TimerTask that checks if some of the first n% (20% in the moment) are active for given time (5 seconds in the moment, and the task checks on 4 seconds). If there are such processes, they are unregistered - stopped and removed from the registry.
  • The NativeBridge interface will have public long getLastCalled method which will be used to sort the bridges in the registry. A TreeSet will be used for the sorting.
  • When a NativeBridge's stdin or stdout is used the last called time will be updated to be the current time.
  • Also the asserts for running will be removed, because the processes now can be stopped, but always when reading or writing to the natives they will be started if stopped.
  • The registry will be singleton for now, if you have problems with that give suggestions (The prolib registry is also a singleton)
• Quiting Sophie 2:
  • In the MainWindowLogic#ON_CLOSE the line MediaUtil.getDefaultAudioOutput().stop(); will be changed to NativeBridgeRegistry.get().unregisterAll();, which unregisters all the bridges in the registry, this works (tested).
  • I recommend testing on windows too.

[meddle]

Design related code

[meddle]

Found problem:

• The Browser process when restarted does not resize and load the Browser content properly:
  • Save the size and the last loaded url of the browser on resizing and loading and when restarting the process resize it and load the saved url.
  • Make sure that every time the timeline of the page is restarted the browser shows it initial set url.

[pap]

• It would be nice if in the JavaDoc you mention the measurement units (seconds, miliseconds, etc.)
• Generally it is good to have some reason when you chose these constants you use.
• I think that it may be better if you put the class specific things first and after them the nested classes.
• Why don't you name "unregister" "unregisterFirst" ?
• Otherwise seems ok, provided that there will not be too many processes to manipulate.
• It may be good if you mention somewhere in the javadoc (perhaps that of NativeBridge) that processes get killed and that natives themselves should be designed stateless enough (The browser reminded me about that).

[meddle]

Implementation code

[meddle]

Implemented and tested, with the design comments in mind. Test it on the Mac I presume, but may be windows is good too to try if the ugly windows exception is gone...

[pap]

• Commited in [8569].
• You could have a bit better variable naming :)
• I like it so much

[meddle]

Closing all the tickets before M Y1