|
RR Crash every few minutes
John Davis [71 posts] |
10 year
|
Both RR 2.61.8 and 2.61.9 crash every few minutes, no matter what I am doing - running, editing, ignoring...
Please suggest what crime scene information I need to collect so that you can catch the bug.
The system is Windows 7 on a RoboMagellan platform with the same hardware (Zotac PC, camera, GPS, ESC, servo controller) that I have been using for years with RR. It works fine, but crashes so frequently now that it is unusable for competition.
|
|
|
|
Steven Gentner from United States [1446 posts] |
10 year
|
It should be asking you when you run it next to send us the error report. That's what we need to see what might be going on.
If you are not getting these (i.e. a crash, restart and NO prompt to send us an error report) then you probably have a crash outside of RR. Most likely in the camera driver. Did you recently purchase a new camera?
STeven.
|
|
|
|
John Davis [71 posts] |
10 year
|
Same old Logitech QuickCam Pro 9000. The only prompt offers me the choice of how I want to fail. I can "Check online for a solution..." or just close. The attached JPG file shows the beginning of the detailed log, but it doesn't look like anything you would write.
|
|
|
|
Steven Gentner from United States [1446 posts] |
10 year
|
John,
Can you try right clicking and select "Run as Administor" on RoboRealm.exe, rut it and then let it crash it should create a
c:\RoboRealm_Crash.log
file on your root folder. Can you send that to us? Or run the app again and it will ask to send it our way. That will contain more information about what may be happening.
The log image you attached does help a little but its error location doesn't seem to make much sense.
Thanks,
STeven.
|
|
|
|
John Davis [71 posts] |
10 year
|
I directly ran the the actual RR executable, not the shortcuts, as administrator. RR crashed, but still no crash report. :(
Windows crash dialog image is attached.
If I select the option of verbose log, RR does create a mind-numbing list of ordinary events called RoboRealm.log.
BTW, I monitored the virtual memory size and didn't see any sign of memory leaks.
It looks like RR is getting a null data value where it expects something else. I'd be happy to run a debug version that would trap errors instead of just dying.
|
|
|
|
Steven Gentner from United States [1446 posts] |
10 year
|
John,
That log file would probably provide some insight ... I sent you a direct email so that you can include it in a reply along with your robofile. That will help to track down this issue.
Thanks,
STeven.
|
|
|
|
John Davis [71 posts] |
10 year
|
I sent you the renamed old log via direct email along with the program. As noted, even though I switched verbose mode back on, RR did not create a new RoboRealm.log.
Is this logfile created once and for all at install time? Can I create a new one with an editor? Should I undo the rename?
|
|
|
|
Steven Gentner from United States [1446 posts] |
10 year
|
The logfile should be generated each time you turn on the verbose mode. Most likely, the permissions prevented you from doing so. When this happens RR will try to save that file elsewhere in a virtual folder depending on what Windows version you are running. It might save it on your desktop, %USERPROFILE%, or in the RoboRealm folder itself (c:\Program Files\RoboRealm) or %appdata%\RoboRealm or %localappdata%\VirtualStore
Note, we obviously wound't make this so undertermined, it is windows that is causing the file to be stored in these potential locations. Can you take a quick look in these locations to see if you see those log files?
STeven.
|
|
|
|
John Davis [71 posts] |
10 year
|
VirtualStore had a small (122 byte) non-ASCII file from April 2013 called RoboRealm.log. It also had a text file, working.robo, from yesterday.
However, I did find RR .dmp files in C:\Users\John\AppData\Local\CrashDumps. The most recent ones are 6.41 MB, so I will use direct email, starting with the latest one.
|
|
|
|
John Davis [71 posts] |
10 year
|
Update: I made sure I was searching the entire hard drive and verified that RR is no longer creating any .log files anywhere, with or without admin privilege.
(The search capability of Windows 7 is a well-known embarrassment for Microsoft; I I used both MS and Agent Ransack to scour my one and only drive.)
I ran a bare program with only the camera module and couldn't get it to crash. There is some Python code in my real program but it resides inside an if-else block that runs only in autonomous mode. The program crashes after startup in the default manual mode.
RR knows where to save programs - my Documents folder; can it write crash logs there by default?
|
|
|
|
Steven Gentner from United States [1446 posts] |
10 year
|
John,
Thanks for the debug file ... we're still trying to understand that. As we can't replicate the error what you can try doing is disable all modules (similar to what you did with testing the camera without any modules) and reenable them until you start getting crashes. Since RR is not creating any crash files the error is most likely outside of RR. So, yes the python module is a possibility but not in your case based on the startup conditional. Other modules like the GPS, Maestro, etc. also have outside drivers that RR interfaces with ... they are very likely candidates of the problem.
If you can't get RR even started with that robofile, you can edit it directly using a text edit and add the word 'disabled' without the quotes as in
<tagname disabled>
and load that version.
We'll continue to see what we can find but this process is much easier on your side since you can replicate the error.
Thanks,
STeven.
|
|
|
|
John Davis [71 posts] |
10 year
|
Thanks for continuing excellent support. I have started disabling up the chain as follows: first, disable the module(s) while RR is running; second, save the program, overwriting the program of the same name; finally, exit RR and restart. RR options are set to restart that known file name.
If the program dies, I leave the disabled module still disabled and disable the next module(s) upstream. If the program doesn't die after a long wait, I enable that module again and disable the next module(s) upstream.
So far, the COG module is the only one that has been re-enabled I just now disabled the Button Control module and saw RR die in the usual way. What was odd was that the Button Control panel popped up even after I restarted (and, yes, the module was still shown as disabled). Next to be disabled is Blob_Replace.
|
|
|
|
John Davis [71 posts] |
10 year
|
Finally narrowed it down to one (and only one) module, Serial. In my program, this module uses a virtual com port in the Pololu Micro Maestro to talk to a Sure magnetic compass with ASCII strings.
The module works perfectly when the program isn't crashing and I haven't modified the hardware or the Serial module for about a year.
I'll check for an updated driver on the Pololu website. I am uploading the program as is, with everything else disabled.
 program.robo
|
|
|
|
John Davis [71 posts] |
10 year
|
I was using Pololu maestro-windows-121203 which is about a year old. After installing the latest version, maestro-windows-130422, RR still crashes with only the Serial module active.
I don't know what else to try.
|
|
|
|
Steven Gentner from United States [1446 posts] |
10 year
|
John,
Thanks for the debugging. Narrowing it down to one module really helps.
We are attempting to replicate the error but need a better idea on what output the board is sending. Is it something like
Angle=150deg
STeven.
|
|
|
|
Anonymous |
10 year
|
Yes. Whatever is in the number part of the string is put into the Heading variable.
I suppose once in a while the compass sends a bad value. Is your parser robust enough to handle arbitrarily malformed strings?
|
|
|
|
Steven Gentner from United States [1446 posts] |
10 year
|
Yes, that was what I was wondering.
We did find a problem that probably caused your issue. Setting the console to refresh every couple seconds causes a buffering of the information coming back from the board. This buffer was overflowing by 1 byte (that's all it takes) which was very effectively causing the application to crash WITHOUT a crash report. So your machine was probably able to write the crash report but because the crash was so strategic it was unable to even provide that report.
If you download the latest version this should hopefully be resolved.
Thanks,
STeven.
|
|
|
|
John Davis [71 posts] |
10 year
|
|
Looks like you fixed it. With RR 2.61.11, my RoboTruck program has been running fine for a few hours now with all modules re-enabled. Thanks.
|
|
|
|
Steven Gentner from United States [1446 posts] |
10 year
|
Thanks for letting us know it worked.
Cheers,
STeven.
|
|
This Thread
