Fix error "no known conversion for argument 1 from ‘sqlite3_int64
{aka long long int}’ to ‘PositionedID& {aka long int&}’"
(Both those types are int64, but that evidently isn't enough)
This makes the behaviour when FGRun (or other GUI launchers)
send bad options much more explicable. Non Windows/Mac systems
get an SG_ALERT on the console as before.
Firstly search model in $fg-root/AI/ path,
if not found, search model in --fg-aircraft paths,
if not found, search model in $fg-root/Aircraft path
This commit should solve the complaint "I have an aircraft in a separate Aircraft/ folder
but during MP session other user are shown with the aircraft from the $fg-root/Aircraft path
instead of this behaviour I want to see the aircraft from my separate Aircraft/ folder"
By Gijs with additions by James. Menubar is disabled at present
since it's not usable and potentially inferior to other solutions,
but committing the code so it doesn't rot.
Adapt to corresponding SG change to make SGMaterialLib be ref-counted,
and have the 'reload-materials' command notify the tile-manager of this,
so it can update the options struct it passes to new tiles.
Restarting the GUI could cause Cocoa menus to misbehave, due to
destruction behaviour of SGBinding. Use new clear() helper in the
short term to work around this.
Warn (but do not require, since people would probably complain) when
added paths do not contain an Aircraft subdir. When people specify
a path which ends in /Aircraft, use the parent directory.
In threaded OSG drawing, MapWidget::draw runs in the render
thread context, but touches lots of main thread state. Move most of the
work to an update() helper run on the main thread instead, so draw()
mostly makes pure GL calls.
(This is a fix for 3.0, the real solution is to migrate to a
Canvas-based map and GUI)
On Mac, the OS is mapping shift+wheel to horizontal scrolling for
those of us without a mighty-mouse or trackpad. Detect this case
and map back to vertical scrolling inside FlightGear for the moment.
https://code.google.com/p/flightgear-bugs/issues/detail?id=1286
When tiles are being loaded, re-evaluate the materials cache once
per update cycle. (This is probably too often, but no worse than the
previous approach).
- borrow some code from the new reset codepath to cleanly stop the
OSG pager thread during normal shutdown. Avoids a race-condition
destroying various structures the pager thread might be modifying.
