More info on Monday crashes.

[JimBecker17]

This is in regard to the crashing behavior have noted which seems to occur much more on Monday mornings than other times.

This morning for the first hour eSignal was crashing on me about once every one to two minutes, between 6:30 and 7:30ish (PST). Now for the past half hour, from 7:30 to 8:10 it hasn't crashed at all. But first hour probably 40 total crashes. Makes it a bit hard to watch the market, eh??

I ran my sample test app which submitted to Robi, and actually it didn't cause crashing any faster than normal. It doesn't do DDE quotes, which might be the difference.

The error is similiar to the one that kept getting in the pre-7.7 releases, but what looks to be the null pointer offset is 0x14 rather than the previous 0x28.

The text of the message is:

"The instruction at "0x5ad71531" referenced memory at "0x00000014". The memory could not be "read".

The attached BMP shows one of the crash dialogs, coming in "DBC Tick Message Sink". The location where it tags this can vary.

Would be great to see this resolved!

-Jim




Woops - spoke too soon. 8:14 and crashed again. The crash does happen intermittently outside of the Monday window, probably ten to twenty times during the day, but whatever is happening on Monday morning, probably server side generated data, is what pushes it much more than other times. Can run fine for an hour then get a crash, when not actually doing anything programmatically.... Hope this helps find this!

[DuaneG]

Hi Jim,

It doesn't look like the image you included made it through. I see that it was a .JPG which could explain it. You may want to try using a .GIF or .PNG file instead. Sometimes, the file size limitation could be an issue.

[JimBecker17]

Let's see if can attach one more time.

[Guest]

Jim,

Can you pinpoint which function is crashing if you run it in debug mode? What function are you calling? Which symbol(s) is it crashing on? I need to know more about what you are doing in your application to determine what needs to be fixed.

[JimBecker17]

Robi,

The crashes are asyncronous to what my logic is doing. My logic does the same thing of using DDE for current quotes and RequestHistory for 5 minute intervals, and also setting the current linking symbol. There is no T&S going on at this point. Don't believe it's related to a specific symbol or anything like that.

The test program I sent you does these calls, with the exception of the DDE quotes, and will always crash eSignal in 4-8 minutes. But I'm not sure it's the same crash going on.

My software will run just fine for long periods of time, but eSignal will crash at irregular intervals with the same error, which appears to be a null pointer reference. This existed in a slightly different form in the pre-7.7 releases as well.

The crash is always from the same address, so if there is a map file from the build hanging around someone could pin down where it's happening.

If you have an eSignal with debug info in it I should be able to run this and attach the debugger to find out more specific info on call stack and data at time of crash.

You can email this to me if desired. I would be happy to do what can be done to solve this.

-Jim

[JimBecker17]

Robi,

Just reviewed one of the old threads about this problem:

http://forum.esignal.com/showthread.php?s=&threadid=9867&highlight=isentitled

Remember now that DDE wasn't a factor, that kept running.

If you see in the middle of this thread the gentleman describes how he is seeing the crashes as well - mostly first and last hours of the day. 7.7 a bit worse than 7.6.

Chris James and I have been talking about this for a while now on some other threads, and on the phone. He was seeing it faster on dual Xeon machines, and actually keeping track of times it occurred.

-Jim

[JimBecker17]

Tuesday followup.

This morning the market has actually been more active than yesterday, and there are no changes in my software or the eSignal software.

So far as of 8:00 have only had one crash of eSignal, at 7:14.

This sure seems to lead one to the conclusion the server side of eSignal can generate data which causes the crashes.

Mondays aren't any different for me than other days, but in the past few months of seeing/reporting this problem suggests it might have something to do with eSignal server farm load...

-Jim

[Guest]

Well, Mondays are actually typically ligher than Tuesdays and Wednesdays (must be because of the weekend). Load was not unusually heavy. I don't think it is data on the server that is causing the problem, but something with the function that is called. The null pointer theory is a good one.

Is there a specific function you are calling that is causing the error? It needs to be tied to some function call.

You are using 7.7, right? Build 679?

[JimBecker17]

Yes, am using the latest build. But this problem has been noted for at least four months I believe. It was in 7.6 and 7.7 betas.

The change with the 7.7 release is at least the offset changed from 0x28 to 0x14. There was a thread about the original crash from July:

http://forum.esignalcentral.com/showthread.php?s=&threadid=9291&highlight=0x00000028

In this I believe one crash was found, but the one I've been seeing was not found.

I continue to get crashes also when exiting the app, but the ones when running the app are the bummer.

Not to point fingers or whine too much, but some of the data which I get through RequestHistory sometimes has errors. I haven't tracked down if it's something my code might be doing and the data is actually good or the data I'm getting from the servers is corrupt. This could be my code, but it might also point to bad data from the server. Or one symbols data gets mixed with another symbol. Normally I clean and recollect data periodically if it looks bad, but do see it happening.

If you need to tie it to a function call, it would probably be RequestHistory. That's what I do the most, always 5 minute and in multiples of 90. One multiple per day back I need to go, which is normally not more than six.

--

As of 10:20 there was another crash at 8:14 after the first one at 7:14, but no other crashes. Don't know what difference there could be between Mondays and other days, but for three-four months now have noted there is this propensity to crash on Mondays. Perhaps your servers are singing the "I don't like Mondayz" blues...

-Jim