Troubleshooting¶
The log¶
With starten.sh it is here:
less "$FG_BASIS/home/fgfs.log"
Typical symptoms¶
What you see |
Possible cause |
|---|---|
New textures don’t load |
The XML is only a link in the overlay, see Overlay |
Black hole at the airport |
Mixed tile sizes, log says |
No aerial imagery |
Tiles in the wrong folder. The log must show |
Arctic on the equator |
DDS not mirrored |
Trees in the middle of a field |
Random trees on, see Weather |
Snow in summer |
Snow line not set, see Weather |
Green lakes or green sea |
See Green water |
Script ends with |
FlightGear stalled too long, increase the wait |
Crash in the graphics driver¶
The whole search is in the article Absturz auf AMD bis in den Treiber verfolgt (German).
The top readable line in the stack is not automatically the cause. Debug symbols show more. Debian provides them through debuginfod.
coredumpctl list fgfs
coredumpctl info
Take the library and addresses from the output. With the build ID you fetch the symbols and resolve the addresses:
readelf -n /usr/lib/x86_64-linux-gnu/libgallium-*.so | grep "Build ID"
curl -o gallium.debug https://debuginfod.debian.net/buildid/BUILD_ID/debuginfo
addr2line -f -e gallium.debug 0xADDRESS
“command stream overflowed”¶
On Mesa 26.1 with a Radeon I saw amdgpu: command stream overflowed while loading, sometimes in long series. The overflow was in si_draw_vertex_state. As a precaution you can try:
export OSG_VERTEX_BUFFER_HINT=VERTEX_BUFFER_OBJECT
Warning
This is not a proven fix. Later it also ran stable for me without the variable, and I could not trigger the crash again. With a bug like this, always test the control case without the change as well.