david [Wed, 1 May 2002 23:13:36 +0000 (23:13 +0000)]
Regularized the position initialization:
1. If /position/longitude-deg and /position/latitude-deg are in range,
use them.
2. Otherwise, if /sim/startup/airport-id is not empty, use it.
3. Otherwise, set the lon/lat to the middle of the KSFO field.
The default used to be Globe, AZ, but that doesn't make sense since we
don't distribute that scenery by default any more.
With this change, starting from a save file seems to work properly:
Bernie Bright:
- Added initial_value argument, in milliseconds, that specifies when
event is first run. The default value of -1 triggers the event
immediately as per the existing behaviour. A repeat_value greater than
zero runs the event no less than every 'repeat_value' milliseconds
afterwards. A repeat_value of zero deletes the event.
- Modified run queue behaviour such that only one event per frame is
run.
david [Thu, 25 Apr 2002 16:23:16 +0000 (16:23 +0000)]
UFO FDM patches from Jonathan Polley:
I have finished working the bugs out of my "Enhanced" version of the Magic
Carpet FDM. The UFO FDM works the same as the Magic Carpet with the
following exceptions:
The aircraft's pitch is determined by the Elevator setting (with a pitch
rate of 45 degrees per second).
Climb is forward velocity * sin(pitch)
Speed is forward velocity * soc(pitch)
Roll is tied to the aileron (again, with a roll rate of 45 degrees per
second)
Turn rate is sin(roll) * 45 degrees.
Yes, this does mean that turning is quite a bit more sluggish than
climbing.
If you are wondering why a UFO FDM, it is because it's best that I not
replace the existing magic carpet and this FDM does behave like a UFO (it
hovers, can spin and pitch while hovering, and does not obey the laws of
physics ;)
There was an integer overflow in the way elapsed_time_ms. Because the
SGTimeStamp "-" operator returns it's result in usec's, there is an upper bound of 37.8 minutes on the maximum difference this operator can reliably report
before the math overflows. I change the global "int elapsed_time_ms" variable
to a "double sim_time_ms" and restructured how the value is calculated.
----
The practical result of the overflow bug is that a large negative dt was
passed to the event manager and the end effect was that no events would
be run until their counters caught up ... in another 37.8 minutes or so.
Detect and recover from a math overflow that can possibly prevent panel text
from updating if the instrument panel is not drawn for more then 35.7
minutes.
Tracked down a potential segfault when trying to audibly ident a vor station.
This led to an investigation into why the ident playing didn't work anymore.
Recent changes in the sound manager broke some assumptions the radiostack
code was making. These patches should hopefully fix all that back up.
Erik, please review these changes to double check I didn't do more damage
than good. :-)
Jim Wilson:
These are the updates for the View manager properties. Removed the last of
items (within the viewer/viewmgr) hard coded to view number. Added support
for per view configuration of ground level nearplane value. Tower views look
very nice with little or no z-buffer problem in the models. Pilot offset
dialog can be used to move eye in all views.
Fixed a recently introduced bug which cause the ground lighting to be
wrongly "lifted" a very, very, very large amount more than it was supposed
to be. I guess the moral of the story is what goes up, must come down. :-)
david [Mon, 15 Apr 2002 15:28:11 +0000 (15:28 +0000)]
This module was filtering out all stations with localizer offset
>90deg, even though src/Cockpit/radiostack.cxx has code to deal with
those; as a result, no backcourses were getting through.
This fix allows the common case of backcourses to work again, but
breaks the uncommon case of a runway using the same frequency for two
separate localizers. That special case will have to be detected
somehow. Still, this fixes more approaches than it breaks.
david [Sat, 13 Apr 2002 21:36:22 +0000 (21:36 +0000)]
Fix for vanishing-model problem: models are drawn in the same scene
graph as the terrain, except for internal cockpit view. The SSG
scene-graph variables (except for the lighting root -- I'll get that
later) are now held in globals.hxx.
FGModelMgr::draw() is obsolete; I'll remove it in a future revision.
david [Sat, 13 Apr 2002 13:11:34 +0000 (13:11 +0000)]
Patch from Jim Wilson:
Minor fixes. Made some changes to get lighting correct for time of day
for now. MSVC compatibility fix and returned to clearing z-buffer only in
LOOKFROM (formerly pilot view).
[dpm: removed the last one, since I had already made a similar patch]
david [Sat, 13 Apr 2002 12:11:27 +0000 (12:11 +0000)]
Ensure that the aircraft model is in the same clip plane as everything
else and *doesn't* clear the z-buffer when in external view; tighten
the far plane for internal view back to 100m to give us maximum
precision (this will be important for some instruments; I might even
move it to 50m).
This patch includes the FGLocation class, a few fixes, cleanup in viewer code.
Synced to CVS 19:36 EDT 2002-04-10 (after this evenings JSMsim and Base
package updates).
Description:
Added FGLocation class which is new home for calculating matrix rotations.
Viewer can now be configured to access rotations created by the model rather
than repeating the same calculations again.
Changed model initialization for the time being so that its location data is
available for the viewer (currently required by other subsystems). At some
point we can move this back to fg_init along with the viewer initialization.
Seperated the update from the draw function in the model code. The viewer
code needs the same matrix data, and moving the update portion at this time
does not increase the number of matrix math iterations.
Moved the model draw so that it always appears "in front" of lights and clouds.
Reogranized viewer update routine for using the FGLocation class and
simplified some more tasks. The routine is fairly easy to follow now, with
the steps ordered and spelled out in comments.
Viewmgr only updates the current (visible) view now, with the exception of an
old reference to "chase view" that will be corrected in forthcoming changes.
Also will be doing some work on the viewmgr outputs.
Model is now clears the z-buffer in all modes. This will be changed with the
next viewmgr update. The only side effect is that models always disappear
when over 5km distant from the eye point (can't really see them anyway:-)).
Other than a flag to indicate "internal" view I don't anticipate the
configuration interface for viewmgr/views will be changed a lot for now. It
is close to done. The next viewmgr update will however rework the outputs so
may change location.
This code will run with the previous version of preferences.xml, but will run
faster with the newer version. I am attaching a preferences.xml that should
not be commited before the code. All the changes are in the /sim/view section
and should show a simpler view configuration that references model locations.
Note that I've added a 2nd tower view in "lookfrom" mode for illustration
purposes. You can look around using the mouse. You may want to remove that or
comment it out.
This patch includes the FGLocation class, a few fixes, cleanup in viewer code.
Synced to CVS 19:36 EDT 2002-04-10 (after this evenings JSMsim and Base
package updates).
Description:
Added FGLocation class which is new home for calculating matrix rotations.
Viewer can now be configured to access rotations created by the model rather
than repeating the same calculations again.
Changed model initialization for the time being so that its location data is
available for the viewer (currently required by other subsystems). At some
point we can move this back to fg_init along with the viewer initialization.
Seperated the update from the draw function in the model code. The viewer
code needs the same matrix data, and moving the update portion at this time
does not increase the number of matrix math iterations.
Moved the model draw so that it always appears "in front" of lights and clouds.
Reogranized viewer update routine for using the FGLocation class and
simplified some more tasks. The routine is fairly easy to follow now, with
the steps ordered and spelled out in comments.
Viewmgr only updates the current (visible) view now, with the exception of an
old reference to "chase view" that will be corrected in forthcoming changes.
Also will be doing some work on the viewmgr outputs.
Model is now clears the z-buffer in all modes. This will be changed with the
next viewmgr update. The only side effect is that models always disappear
when over 5km distant from the eye point (can't really see them anyway:-)).
Other than a flag to indicate "internal" view I don't anticipate the
configuration interface for viewmgr/views will be changed a lot for now. It
is close to done. The next viewmgr update will however rework the outputs so
you can expect that the data that viewmgr is writing to the property tree
may change location.
This code will run with the previous version of preferences.xml, but will run
faster with the newer version. I am attaching a preferences.xml that should
not be commited before the code. All the changes are in the /sim/view section
and should show a simpler view configuration that references model locations.
Note that I've added a 2nd tower view in "lookfrom" mode for illustration
purposes. You can look around using the mouse. You may want to remove that or
comment it out.
david [Tue, 9 Apr 2002 18:58:24 +0000 (18:58 +0000)]
Removed current_model global variable and added a reference in
FGGlobals instead (we still need to do the same with current_input and
a few other globals).
andy [Tue, 9 Apr 2002 04:55:16 +0000 (04:55 +0000)]
Change the idle MP to reduce idle power a bit. This is basically a
hack, and should be replaced with a sane idle model that has some
physical justification.
david [Sun, 7 Apr 2002 21:20:16 +0000 (21:20 +0000)]
Patch from Frederic Bouvier (FIXME comments added by David Megginson):
I made these modifications to correct the problem of getter functions
returning pointers to local variables. Those variables are no longer
locals and become static. One should be aware that it kills reentrancy
!
david [Sun, 7 Apr 2002 21:13:56 +0000 (21:13 +0000)]
Patch from Martin Dressler:
This patch moves built-in Class (for now only mag-ribbon) into special
directory as you have written it in TODO: in comments of this class in
panel_io.cxx. IMHO it is good idea. I want to play with built-in
classes and OpenGC and this will be useful.
david [Sun, 7 Apr 2002 15:04:49 +0000 (15:04 +0000)]
Patch from Jim Wilson:
Made a couple more adjustments, now the function builds the
transformation matrix (sans the translation) directly from the five
inputs. After realizing that it may be necessary for the new location
class to publish a world up vector, comments were added indicating
which values constitute world up.
david [Sat, 6 Apr 2002 17:23:18 +0000 (17:23 +0000)]
Viewer update from Jim Wilson:
Minor patches to initial tower view code. Added calculation for "tower" at
any airport, placed a couple fractions of a degree off the center. Moved the
farplane from 100m to 5km so that "lookfrom" mode can see the model. 5km is
still in a reasonable range for a depth buffer. Note that looking at a model
anything but a small distance will cause some problems on older cards.
Temporarily changed viewer to always report elevation/lon/lat of the aircraft
in order to avoid a problem with the ground trim getting recalculated when
tower and aircraft are on different tiles.
Here is the 'boostified' event manager plus relevant changes to
radiostack.cxx, fg_init.cxx and main.cxx. If these changes are accepted
then you can remove Time/event.[ch]xx and Include/fg_callback.hxx from
the repository.
david [Fri, 5 Apr 2002 18:49:04 +0000 (18:49 +0000)]
Viewer update from Jim Wilson:
Tower View and viewer config is in place. Note that the interface is
still in a state of flux. A couple of the config items (namely the
offsets) are still using the old settings. The tower is hard coded
into the base package for a position off the starting runway at KSFO
and is probably not in the right place for there even. Looks pretty
cool though! Tower View is the third view. If you aren't at KSFO
you'll just see blank space in view 3. It's looking through the earth or
something like that :-). Important note: zoom in with a few
hits of the "x" key to see the plane better in tower view.
david [Fri, 5 Apr 2002 18:48:04 +0000 (18:48 +0000)]
Patch from Jim Wilson to realign model slightly in new model code:
This should fix it. The problem wasn't the rotations but a difference in
the translations. Note that this fix also puts the wheels back on top of the
pavement when in external view.
david [Fri, 5 Apr 2002 18:46:47 +0000 (18:46 +0000)]
Removed dependency on viewmgr.hxx from globals.hxx, so that changes to
the viewer code do not require most of FlightGear to be rebuilt.
Added some now-required includes to environment.cxx.
david [Fri, 5 Apr 2002 16:55:30 +0000 (16:55 +0000)]
JSBSim fix:
Changed the steering gain from +/-0.1 to +/-1.0, so that the steering
angle for the nosewheel (when present) is the same as the rudder
deflection angle. That's probably not exactly right, but it's much
better than we had before -- you can now steer the plane on the ground
reasonably during taxiing.
david [Fri, 5 Apr 2002 03:19:34 +0000 (03:19 +0000)]
Moved FGAircraftModel subsystem out into its own directory, and
separated out a new class, FG3DModel, that allows any arbitrary 3D
model to be positioned, oriented, and animated (not just the
aircraft).
david [Thu, 4 Apr 2002 17:51:40 +0000 (17:51 +0000)]
Let exceptions fall through to main, and use cerr to report them (in
case logging is disabled). This way, when people specify a
non-existant aircraft or have an error in a custom XML file, they'll
get an error report, at least.
Attached are a fairly extensive series of patches to the ATC
system. A chap from Germany called Alexander Kappes (cc'd) got
in touch with me a few weeks ago and has written the start of
Approach control. At the moment tuning in to a valid approach
frequency (Dortmund or East Midlands) should result in vectors to
a spot about 3 miles from the active runway, and a telling off if you
stray too far from the correct course, in the console window. He
seems to know what he's doing so expect this to improve rapidly!!
I've added a rudimentry AI manager and a hardwired Cessna at
KEMT on the runway - I'll remove it before the next release if I don't
have it flying by then. There seems to be an issue with framerate
which drops alarmingly when looking at it - I've a feeling that I've
possibly created several Cessnas on top of each other, but am not
sure.