Changelog 3.4: Difference between revisions

m
→‎Major enhancements in this release: disclaimer: these quotes are just kept as reminders here, so that we can grow a list of changes over time ...
(Created page with "{{draft|changelog|This changelog is currently being written for the FlightGear 3.4 release. To see what is being worked on, check out Changelog 3.4. Feel free to help! If...")
 
m (→‎Major enhancements in this release: disclaimer: these quotes are just kept as reminders here, so that we can grow a list of changes over time ...)
Line 30: Line 30:
'''Core'''
'''Core'''


{{FGCquote
  |I have just pushed a partial fix for the issue to SimGear.<br/>
The event now uses property polling instead of attaching listeners to <br/>
retrieve the values for the uniforms.<br/>
<br/>
Without the change, I was able to reproduce the crash locally within a <br/>
couple of minutes. With this patch,<br/>
I havn't seen a crash over a period of 3.5 hours constantly watching AI <br/>
traffic over KSFO building up.<br/>
<br/>
This only changes the behaviour of scalar properties, arrays et. al. <br/>
have not yet been changed.<br/>
Also, this is still not thread safe - no locks have been implemented. If <br/>
a getValue() for a property gets called<br/>
from within the OSG thread while the main thread updates the same value, <br/>
this might result in a wrong value<br/>
for  multibyte values.However this should be automatically corrected in <br/>
the next frame and not crash the system.
  |{{cite web |url=http://sourceforge.net/p/flightgear/mailman/message/32774953/
    |title=<nowiki>Re: [Flightgear-devel] crash in SGPropertyNode::fireValueChanged</nowiki>
    |author=<nowiki>Torsten Dreyer</nowiki>
    |date=<nowiki>2014-08-29</nowiki>
  }}
}}
{{FGCquote
  |As a side effect, I have noticed that the load for the xml-property rule <br/>
subsystem has decreased on my system<br/>
significantly. This subsystem generates values for most of the <br/>
environment properties responsible of the<br/>
instantiated effects. Without the patch, thousands of listeners were <br/>
triggered each frame resulting in an average<br/>
of 1ms execution time per frame for the proprule subsystem. With the <br/>
patch, this has decreased to an average<br/>
of 0.07ms per frame. This is on a 3.5GHz 6 Core Xeon. Weaker machines <br/>
might benefit even more.
  |{{cite web |url=http://sourceforge.net/p/flightgear/mailman/message/32774953/
    |title=<nowiki>Re: [Flightgear-devel] crash in SGPropertyNode::fireValueChanged</nowiki>
    |author=<nowiki>Torsten Dreyer</nowiki>
    |date=<nowiki>2014-08-29</nowiki>
  }}
}}
'''Aircraft Modeling'''
'''Aircraft Modeling'''