Open alexdawn opened 5 years ago
please provide all necessary input files for reproducing the crash. It would be preferable if you could cut down the scenario to a smaller version which still crashes. This tool might help: https://sumo.dlr.de/wiki/Tools/Routes#cutRoutes.py
also, please check first whether the crash happens with version 1.3.1 as several crashes have been fixed since 1.2.0
I've updated to 1.3.1 and it is still crashing although now at 08:01:21 I'll see if I can still replicate with a reduced network although I haven't come across any issues so far with the smaller model (this is the reduced network with the addition of buffer links and external demand, the smaller model has had no crashes like this)
Problem signature:
Problem Event Name: APPCRASH
Application Name: sumo-gui.exe
Application Version: 0.0.0.0
Application Timestamp: 5d648982
Fault Module Name: sumo-gui.exe
Fault Module Version: 0.0.0.0
Fault Module Timestamp: 5d648982
Exception Code: c0000005
Exception Offset: 0000000000151b60
OS Version: 6.1.7601.2.1.0.256.48
Locale ID: 2057
Additional Information 1: e339
Additional Information 2: e3399d2a3b619755bbea1f1714ea00f9
Additional Information 3: 07fd
Additional Information 4: 07fdb20d0ea47bbbc70c7064615b1bbe
Read our privacy statement online:
http://go.microsoft.com/fwlink/?linkid=104288&clcid=0x0409
If the online privacy statement is not available, please read our privacy statement offline:
C:\Windows\system32\en-US\erofflps.txt
The simulation runs to completion with just the bus routes files loaded. One of the car trips must be generating the crash somehow? Not the number of trips are only ~900 vehicles are loaded when the crash occurs. Not sure if it can be invalid routes or anything as DAROUTER and MAROUTER run fine?
while the clock stops at 28861
, there are some logs from 28862
for lane changes but no logs for floating car observers or netstate, for these two the logs cut halfway through a tag in 28861
and for lane changes halfway through a tag in 28862
<change id="4656.0" type="car" time="28862.00" from=":3900_4_2" to=":3900_4_1" dir="-1" speed="9.49" pos="8.45" reason="strategic|urgent" leaderGap="13.97" leaderSecureGap="10.28" leaderSpeed="9.09" followerGap="None" followerSecureGap="None" followerSpeed="None" origLeaderGap="None" origLeaderSecureGap="None" origLeaderSpeed="None"/>
<change id="4447.0" type="car" time="28862.00" from="4407.flarex_0" to="4407.flarex_1" dir="1" speed="7.16" pos="5.69" reason="strategic|urgent" leaderGap="None" leaderSecureGap="None" leaderSp
aha! @namdre running the model where NETCONVERT has changed to be --opposites.guess false --opposites.guess.fix-lengths false
seems to have fixed the crash, which is a big clue to the problem. I've not seen an issue opposite lane overtaking in the smaller model though potentially some of the buffer links in the larger model are very long and have a very high lane count
The warnings in your first log file are intriguing. If you run netconvert with options --opposites.guess true --opposites.guess.fix-lengths true
then opposite lanes should all have the same lengths. However, the lanes in the warnings are all internal. Please post all options that were used the build the network of the crashing simulation (you can just copy the header from the .net.xml).
Ah I'd grown blind to those warnings, though normally I see them in NETCONVERT not SUMO my arguments were:
netconvert --node-files="$VAGRANT_MODEL_SSH/$1_network.nod.xml"\
--edge-files="$VAGRANT_MODEL_SSH/$1_network.edg.xml"\
--connection-files="$VAGRANT_MODEL_SSH/$1_network.con.xml"\
--output-file="$VAGRANT_MODEL_SSH/$1_network_without_ghosts.net.xml"\
--type-files="$VAGRANT_MODEL_SSH/$1_network.typ.xml" --proj.utm\
--plain.extend-edge-shape\
--lefthand\
--sidewalks.guess false --crossings.guess false\
--default.crossing-width 2.0\
--sidewalks.guess.from-permissions false\
--sidewalks.guess.max-speed 21.9 --sidewalks.guess.min-speed 0.0\
--opposites.guess true --opposites.guess.fix-lengths true\
--junctions.scurve-stretch 1\
--no-turnarounds.except-deadend true\
--log="$VAGRANT_MODEL_SSH/$1_netconvert.log" &&\
netconvert --sumo-net-file="$VAGRANT_MODEL_SSH/$1_network_without_ghosts.net.xml"\
--output-file="$VAGRANT_MODEL_SSH/$1_network.net.xml"\
--edge-files="$VAGRANT_MODEL_SSH/$1_network_patch-ghosts.edg.xml"
Please check whether sumo also warns about Unequal lengths of neigh lane
when you load the network after the first netconvert call.
Looking over the network I've discovered I'd been generating multiple links with the same from
and to
values in this outer area. These duplicate edges might be responsible for the odd behaviour
If you can put together a small failure example I'll be happy to check it out.
On hitting play the simulation crashes 18 seconds in at 08:00:18
The network itself loads fine both in sumo-gui and NETEDIT. It crashes at the same time division in the non gui sumo.exe with the line
StepSegmentation fault= 30.30*RT, ~70030.30UPS, vehicles TOT 2342 ACT 2311 BUF 14)
while it might be related to the demand the initial demand file runs fine in DUAROUTER and DFROUTERFull text in the window:
And the log file upon loading the sumoconfig file into sumo-gui: