I finally found the Nasal memory corruption bug this morning. I was
right that it was triggered by garbage collection, but it wasn't in
the Nasal code. The FGBinding implementation was holding a
SGPropertyNode* and assuming that its callees would never try to use
the SGPropertyNode_ptr interface to refcount it. Nasal did, which
caused the node to be prematurely freed. Trivial fix; horrific
debugging effort. :(
So I've started checking in my Nasal control/input changes. The
keyboard bindings are first, because they're easier to test. I'll get
to the joystick bindings as soon as I can. With one or two
exceptions, nothing should be visibly different to the user. Notable
changes:
+ The v/V bindings for stepping through views are no longer hardcoded
in the XML, and pop up a nice eye-candy label to tell you what view
you have selected. I'll leave it to someone else to come up with a
translation of "Chase View wo yaw" :)
+ Flap steppings are now settable per-aircraft. You can have as many
as you like; the up/down flaps bindings do the right thing. I set
up a sample in the 747-yasim-set.xml configuration. The syntax is
very simple (I have no idea if these values are right, though):
<flaps>
<setting>0.00</setting>
<setting>0.01</setting>
<setting>0.02</setting>
<setting>0.05</setting>
<setting>0.10</setting>
<setting>0.33</setting>
<setting>0.67</setting>
<setting>1.00</setting>
</flaps>
There is compatibility code in place so that aircraft without a
"/sim/flaps[n]" declaration will get the old 0.0/0.33/0.66/1.0
steppings. Note that only the *keyboard* bindings are currently
using this code.
+ Engine settings automatically detect and set properties for the
correct number of engines. No more need to duplicate XML bindings
for a "maximum" number. The XML files are a *lot* smaller for this.
+ The "popup tip label" used for the zoom and (now) view change code
is available for general use as "gui.popupTip('My tip message');".
+ There is a "slewProp()" implementation that handles realtime-based
motion of property values. This will eliminate the framerate
dependencies of trim and view direction changes for joystick
handlers. Note that this is *not* used by the keyboard handlers,
becuase the OS generates its own key repeats that we'll have to turn
off first.
I've tried to test this, but something probably slipped through. Let
me know if I broke something. I'll hopefully have a chance to work on
the joystick bindings tonight.
Andy
_______________________________________________
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel
