How do I debug a no-message crash to desktop?

[KimD]

Livecode 9.0.0 Indy running on Windows 10.

1) In both the IDE and Windows Standalone I'm experiencing a no-message crash to desktop.

2) Through use of extensive logging I've narrowed the point at which my app terminates down to the single line -

Set the dgText[false] of group "DataGrid1" to Empty

3) However, I never experience a crash the 1st time that this line is run. If I repeat EXACTLY the same actions, the crash will occur on the 2nd, or 3rd, or 10th, etc, time that I execute this line of code.

4) My logfile is recording the dgNumberOfLines of group "DataGrid1" immediately before DataGrid1 is set to empty. The dgNumberOfLines is returning the expected result. "12" to be precise

5) The only other clue that I've got, and it's out-rightly obscure, is that my PC's cooling fans fire up when executing the offending handlers. The cooling fans stay silent for all other parts of my app. LC seems to be doing something that requires a lot of ummppphhhh at this point.

I've got a huge amount of logging in place, and my app is executing all of the handlers that I'd expect in the order that I expect. It just gets to "Set the dgText[false] of group "DataGrid1" to Empty", then sometimes dies.

Any tips on how to debug a no-message crash to desktop would be much appreciated.

Thanks

Kim

[FourthWorld]

This may help:
http://fourthworld.net/lc/4wLogger.livecode.zip

It's a logging tool that will dump a rather verbose log to you desktop. After the crash, open the log and you can at least see the last thing that happened before the crash.

Please let us know what you find. LC should never crash, so once we pin this down we can submit a bug report.

Thanks -

[dunbarx]

LC should never crash

Hypercard never crashed. LC is pretty good, but I have had crashes here and there forever.

I have a stack right now that has been running fine for years, and it all of a sudden crashes LC if I try to open the SE for any control. Otherwise it works normally. I copied the two cards in that stack into a brand new stack, essentially recreating it on the fly. The new stack is solid. I threw away the old one.

Craig

[KimD]

Hi Richard and Craig

Thanks for replying. I agree that crashes to desktop are very rare in LC. This is the only repeatable CTD that I've had in the past 12 months or so.

I've trimmed my stack down to the still-crashing minimum. This is attached. This stack has crashed to desktop (from the IDE) about 5 times for me in the past 30 minutes. I don't know why, but sometimes it takes ages to get the same stack to CTD, but once it starts it then happens all of the time. Very odd.

I haven't tried this stack as a standalone, but its parent crashes to desktop as both IDE and standalone.

I've also attached the 4W logfile for a crash-to-desktop that I had with this stack. The last messaged logged is -
_table.DeleteDataControls

I can easily re-code to avoid using the problematic Set the dgText command. With the benefit of hindsight, after a cancel operation the contents of the data grid should not change, so I don't need to repopulate the datagrid. DOH! But, if my incompetence has uncovered a bug in LC then it would be good to have both fixed.

Regards

Kim

[FourthWorld]

Thanks for that info, Kim.

Is this issue reproducible with the latest build, v9.0.4rc3?
https://downloads.livecode.com/livecode/

They've been doing some work with memory management for that build, so if we're lucky the cause of this issue has already been addressed.

[bangkok]

Yes, it crashes.

LC Indy 8.1.10
and
LC Indy 9.0.4 rc 2

Windows 8.1

Your code... is rather strange.

What's the point to have on the card "update" a preopencard hander.

And :

Code: Select all

go cd "autoreturn"

and then on card "autoreturn" to have :

Code: Select all

On PreOpenCard
    Logevent "Always", ggLogfile, "Card AutoReturn > PreOpenCard"
      Go to card "Update"
 end PreOpenCard

[LiveCode_Panos]

Hello all,

We can reproduce the crash using the sample stack.

Filed this report so as we can track progress on this issue:

https://quality.livecode.com/show_bug.cgi?id=22044

Regards,
Panos
--

[KimD]

Thanks all

My real stack is now working. It was only ever during a "cancel" operation that I was experiencing a CTD, and I re-wrote the cancel handler to avoid the un-necessary "Set the dgText". I haven't seen a CTD since.

Yes - the demo stack is "rather strange". This comes down to (1) this demo stack contains only 5% of the code in my "real" stack, and (2) pilot error. The short version is that, on a change to the source data for the data grid, my script re-opens the card instead of just re-populating the data grid. "Strange" - yes. Should result in a crash to desktop - probably not.

I wrote log files to ensure that I wasn't getting into an unexpected loop and I wasn't trying to perform a data grid operation while I was on "the wrong card". My interpretation of the logs is that my script is OK on both of these counts. Something else is going wrong. It seems to me that this might identify a specific area where the DG stack could be made more robust, but I don't have the skills to take this any further.

Thanks for the 4W logger. Nice!

Regards

Kim

[FourthWorld]

Hello all,

We can reproduce the crash using the sample stack.

Filed this report so as we can track progress on this issue:

https://quality.livecode.com/show_bug.cgi?id=22044

Regards,
Panos
--

Thanks for filing that report, Panos. I look forward to learning the cause.