Step back with a bang?

Hi all,

I seem to keep running into these weird glitches. This has happened to me several times now on different routes. Here are two plots of the black box data from the pertinent section indicating the problem. Both plots are from the same run. The first is time-based and the second is distance-based. At first glance I couldn't find anything out of the ordinary in the route file, and I've been driving the route before without any problems. My computer hasn't exhibited any problems or stuttering.

The interesting part is the Z-shaped glitch. According to the black box data, the train actually jumps backward about ten meters, whilst the emergency brake engages. Also the black box indicates the security device drops the reverser to neutral. On screen this shows up as a violent swing or jump accompanied with a bang and the brake going into emergency. I can continue as soon as the brake disengages. In the review screen it says "Doors opened", although the door interlock light stays on, and the door indicator remains off. If I actually open the doors at speed, it simply leads to a normal emergency brake application.

The funny thing is that the time doesn't match up in the black box data: the security system's reverser position actually goes backward in time, or attempts to be both forward and neutral at the same time, I don't know which. Also the acceleration data goes back in time, i.e. there are two or three acceleration values at some points. 

Has anybody else run across anything like this? Or is my computer going bonkers? Or am I? 

Time-based plot of an example event (Birmingham Cross-City South, between Redditch and Alvechurch):

Distance-based plot of the same event:


I have a few things I can possibly try, I'll report back if I find anything. This seems to be pretty repeatable.

And again; what I did was turn off collisions and reboot my computer. It happened nevertheless. This time the route is Network West Midlands 2014, the 06:55 Norton CS-Hobbs Cross P3 ECS with a 323. And again, as always, nothing on the line, nothing in the route file as far as I could see, no alarms from the security system. There was a loud bang , the train tilted forcefully to the left (sometimes it tilts to the right, sometimes just jumps back a bit), the brake went to emergency, and released as soon as the train had come to a halt. The speed was well within the limit (70 mph), and the train was approaching a double yellow signal, with the AWS alert duly acknowledged.

Now, a brake demand from AWS or TPWS would have shown on the in-cab instrument (have tried that). Opening the doors would simply drop the brake to emergency and turn off the door interlock light (have tried). Operating the DRA would simply disable power (tried that, too).

I went and looked at the track with the outside view - nothing out of the ordinary.

The review screen still says "Doors opened", which I find strange, since they definitely weren't.

These charts shows the actual event as it appears in the black box data. The first chart spans a time of 2 seconds.


The second chart shows the same thing, but related to the distance covered in the 2 seconds, which is 70 metres (vertical lines every 15 m).

Anything curious there in the route file?

No, not as far as I can tell, and I've driven the same route before without any problems whatsoever. I'm beginning to suspect I've made some update on my computer that's not playing nice with OpenBVE.

What version of the Class 323 are you using?

There's basically three there's two on my site, and the Railsimroutes 3D cab one which has a WIP plugin so does not work correctly.

I'm in fact using the Railsimroutes 3D cab train with UkTrainSys.dll version 0.3.1.9. It has worked perfectly thus far. This has happened with a class 58, which I have configured to use UkTrainSys.dll as well, and it has also worked flawlessly. Could the problem be with the plugin? I've been using the same version all along.

I've been using UkTrainSys.dll as a replacement for UKDt.dll, since the former has worked well on Linux thus far and the latter isn't multiplatform.

Hard for me to say as I use Windows of course.

But I don't use it in my releases as it's never worked correctly, that's not surprising as it's an alpha version.

Which a couple of years on from it's release is a bit disappointing, but to be expected as this is very much a in your own time hobby and it's better than nothing if you don't run Windows.

The funny thing is that I've changed basically every UK train to use UkTrainSys.dll because it's worked perfectly thus far...  I can of course try a different plugin to see if something's gone wrong. The versions of mono (the .NET runtime environment), nvidia drivers, OpenAl, and OpenSDL are what has most probably changed in the meantime.

Or else somebody's discovered a latent schrödinbug in UkTrainSys.dll, i.e. found out that it shouldn't have worked in the first place, causing it to stop working for everybody.

I'd say the problem is almost certainly with the plugin, and by the looks of things my guess would be that the AI guard is misbehaving.
An easy test is to delete the plugin entirely; If it's the plugin this will then run fine, but obviously with no safety systems

Try deleting the .Beacon 24 command on the same line as Alvechurch station and see whether that gets you anywhere.
I haven't checked NWM for AI guard support beacons, but I presume they're in there as well.

I'll try and run it in the debugger (Windows mind....) later in the week and see if I can see anything interesting on Windows.

Cheers

Chris Lees

http://www.bvecornwall.co.uk

Thanks, I'll try running with the plugin disabled, vs. the plugin enabled and the .Beacon 24 command deleted and see what happens. I'll report back once I get to it.

Interestingly enough, I disabled the UkTrainSys.dll plugin altogether and ran from Redditch through Alvechurch to the first signal without problems.

Then I ran with the plugin enabled and gathering a debug log until I heard the bang.

The last lines of the debug log indicate that the plugin decides - for some mysterious reason - release the door interlocks, i.e. act as if the doors had been suddenly opened. This happened before a station stop. Here's what happened, the bang is the last three lines:

Code:

01:44:55.4805520 AutomaticWarningSystem Void Prime() AWS: [INFORMATION] - AWS south pole detected at location 10094.7859905906 metres - system is primed and delay period timer is active
01:44:56.5096610 AutomaticWarningSystem Void Update(Double, System.Int32[], System.Text.StringBuilder) AWS: [INFORMATION] - AWS delay period has elapsed at location 10117.7012785543 metres - issuing AWS warning
01:44:57.0376640 AutomaticWarningSystem Void OnAwsWarningAcknowledged(System.EventArgs) Aws: [EVENT] - AWS warning acknowledged at 10128.5388202759 metres
01:46:02.0456180 TrainProtectionWarningSystem Void ArmOss(Int32) TPWS: [INFORMATION] - OSS arming frequency f1 detected at location 11337.7092662947 metres - train speed is 44.2584203870301 km/h, OSS ND armed and timer is started
01:46:03.0634050 TrainProtectionWarningSystem Void Update(Double, System.Int32[], System.Text.StringBuilder) TPWS: [INFORMATION] - OSS ND timer expired - resetting. Train speed is 44.0185529977769 km/h at location 11350.1336126074 metres
01:46:04.6801270 AutomaticWarningSystem Void Prime() AWS: [INFORMATION] - AWS south pole detected at location 11369.8044191808 metres - system is primed and delay period timer is active
01:46:05.7180770 AutomaticWarningSystem Void Update(Double, System.Int32[], System.Text.StringBuilder) AWS: [INFORMATION] - AWS delay period has elapsed at location 11382.5490533051 metres - issuing AWS warning
01:46:06.2519990 AutomaticWarningSystem Void OnAwsWarningAcknowledged(System.EventArgs) Aws: [EVENT] - AWS warning acknowledged at 11388.579642792 metres
01:46:14.9973920 InterlockManager: [INFORMATION] - Interlock brake reset and release request
01:46:14.9975540 InterlockManager: [INFORMATION] - Interlock traction power reset and release request
01:46:14.9977340 InterlockManager: [INFORMATION] - Traction interlock release successful


Here's a normal station stop:

Code:

01:41:51.5747520 AiGuard: [INFORMATION] Calculating new stopping location - this
 will be @ 7936 metres - number of cars specified is 0 (train has 3 cars)
01:41:51.8421400 AutomaticWarningSystem Void Update(Double, System.Int32[], Syst
em.Text.StringBuilder) AWS: [INFORMATION] - AWS delay period has elapsed at loca
tion 7776.54735466075 metres - issuing AWS warning
01:41:52.3443200 AutomaticWarningSystem Void OnAwsWarningAcknowledged(System.Eve
ntArgs) Aws: [EVENT] - AWS warning acknowledged at 7779.57916193623 metres
01:42:28.8208290 AiGuard: [INFORMATION] Stopped successfully within station limits @ 7936.92152105883 metres
01:42:48.9879020 InterlockManager: [INFORMATION] - Interlock brake reset and release request
01:42:48.9897730 InterlockManager: [INFORMATION] - Interlock traction power reset and release request
01:42:48.9900600 InterlockManager: [INFORMATION] - Brake interlock release successful
01:42:48.9920120 InterlockManager: [INFORMATION] - Traction interlock release successful
01:42:51.4257250 AiGuard: [INFORMATION] Ready to start buzzer signal issued


So I disabled the interlock manager (a faulty door interlock would take the unit out of service, at least to my understanding, but this is only a simulation, so never mind), but the bang still occurs. Now it happened when braking into the platform at Longbridge, the last time I was approaching the signal before Longbridge. (B'ham Xcity South again, same route as previously.)

One last thing; I reformatted the Beacon 24 commands to comply with the UkTrainSys plugin manual, i.e. leaving the second parameter blank, since it is always zero... aaaand nope, it still does that thing - this time between Alvechurch and Barnt Green, far away from any Beacon 24's. Next I'll try writing a configuration file for BVEC_ATS, but that can wait until later.

Few points:

Are you using the Ubuntu version of X-City-South, or the version from Anthony's site?

The AI support beacons differ for UKDT/ UKMU designed plugins, and for UKTrainsys.
If you're hitting UKDT beacons, it wouldn't surprise me if the AI is getting a little confused
It shouldn't be opening the doors, but don't quote me on that!

The second paramater of a .beacon command is the object structure index. Zero is fine, I'm not sure what blank is likely do do, other than possibly generate an OpenBVE error log entry.

Final note:
I've had one report from a Japanese user that BVEC_ATS throws a wobbly on Ubuntu and crashes on loading. I can't reproduce this (12.04 VM), but would be interested if you can.

Cheers

Chris Lees

http://www.bvecornwall.co.uk

I can't reproduce this at present using the 323 on Windows.

Ignore the door interlocks- UKTrainsys's interlock manager manages the interlocking between systems that govern what the plugin actually applies in terms of power and brakes.

I don't see anything interesting in your 'faulty' log though-
The last three lines are simply saying that something requested the release of any active interlocks, which was successful.
To be exact, these are the three potential conditions which could cause them:
Reverser moved away from N
An active brake demand triggered by TPWS expires.
An expired deadman's handle brake application.

Can you give me some exact routefiles and locations please

Steve- If you've got specific locations/ bugs also, I'm more than happy to run in the debugger and see if I can get things working right.

Cheers

Chris Lees

http://www.bvecornwall.co.uk

Thanks for the comments.

I'm using the version from Anthony's site, but I've actually changed the beacons to work with UkTrainSys. The problem is that the exact same beacons have worked until now! I'm running Ubuntu 14.04 with all the most recent updates, which has of course updated some supporting libraries, but not the route files or OpenBVE binaries (which are from the official site). This is why I now have an indoor blizzard of dandruff from scratching my head.

I've tested some earlier versions of BVEC_ATS and it's worked fine. I just need to check the panel indices and buttons to come up with a configuration file where I can actually see something happening, then I can check if this works.

Hmm... actually, if this is only a matter of plugins, I have BVEC_ATS already configured on Dexter's excellent 163... I wonder if it fits under the bridges...

...and after a brief trial I have three answers:

1. yes,the 163 does fit
   
2. yes, BVEC_ATS (at least the version I have now, which has worked previously) runs,
   
3. yes, the "bang" still happens.
   

Do you mean the bang happens with BVEC_ATS as well?
Does the doors message appear in the review?

The plugin doesn't touch anything to do with the doors, so this would suggest an OpenBVE bug.
More specifically, I'd guess a recent change in Mono has probably broken something; That'd explain the bug report I've had as well

Now downloading 15.04 to do a proper install rather than VMWare, and I'll see what I can find in the debugger.

Cheers

Chris Lees

http://www.bvecornwall.co.uk

Yes, it happens with BVEC_ATS, and the doors message appears in the review.

No bangs here I'm afraid 

I've just run the 163 through one of the day runs on Birminham X-City South, with no problems other than some semi-intentional atrocious driving. (I was running at 5x speed and not paying too much attention to speed limits)
Specs:
Ubuntu 14.04 (Fresh install, only OpenBVE on here)
Q9400
4gb RAM
Radeon HD6870

Current build of BVEC_ATS also works, much to my relief.....

Any joysticks, gamepads or the like plugged in?
I'd also try with a different keyboard and see where that gets you
Failing that, I'd be inclined to roll a fresh install, as it sounds like something is stuffed.

Edit:
Scratch that, haven't installed the pending system updates....
Will report back once it's updated. (Don't know why the update during install option didn't bother to install them.....)

Cheers

Chris Lees

http://www.bvecornwall.co.uk

I have the exact same joystick and keyboard as before, and they have worked previously.

And yes, everything did work on 14.04 up until this spring, so it must have been some update in the past four months or so that broke something, but I don't know what it is.

Thanks,
JS