Zugg Software :: View topic

Posted: Sat Jul 18, 2009 7:25 am

I have a trigger that seems to be cause of this access violation. It's not something that occurred at all in 3.08, so this is quite annoying.

The trigger is simply:

SubAdmin Joined: 18 Nov 2001 Posts: 5187

It looks like it is connected to the change in handling for the VT100 clear line sequence. Please provide a debug log with the incoming text for a little ways before the AV.

Posted: Sat Jul 18, 2009 3:31 pm

I'll see what I can do. Has happened only once since this post, but I'll start the debug log. Thanks!

Charneus

Posted: Sat Jul 18, 2009 3:44 pm

Ok. Just did #DEBUG test.txt trace.txt and it never created any type of file. Am I missing something here?

Charneus

Edit: Determined it is now #DEBUGFILE instead of just #DEBUG. Yay for reading helps. Will submit the files after the error happens again.

Charneus

Posted: Sat Jul 18, 2009 4:29 pm

Upon further research, I find that it is truly messing up my wordwrap. For whatever reason the access violation is happening, CMUD no longer wordwraps after it, and changing the wordwrap options just creates more AVs. I did do a post on the wordwrap bug earlier, but now I wonder if these two are related.

Charneus

Posted: Sat Jul 18, 2009 8:57 pm

Ok. I've uploaded the test and testraw text files to my website (it would fill up a LOT here.)

The most relevant part is at the bottom, when the access violation occurred. You can find them here:

Test output
Raw output

Hopefully, it'll help. Be mindful of possibly some colorful language if you read through it all. :P

Charneus

Posted: Sun Jul 19, 2009 1:51 am

Just noticed this weird output line on the raw txt:

SubAdmin Joined: 18 Nov 2001 Posts: 5187

Can you narrow it down any further. I didn't see anything in the log that would confirm my initial guess, and couldn't replicate the crash no matter what I tried.

Posted: Sun Jul 19, 2009 2:53 am

Well, the AV happened at the last line. CMUD was closed after that.

I did the debug again, and these were the last lines received when the AV happened again:

Posted: Mon Jul 20, 2009 5:04 pm

Definitely related to the new code that is clearing the line attributes. I couldn't reproduce it with the debug files, but with the crash dump info that you posted I might be able to at least add some exception handling to that part of the code to prevent the crash.

The weird codes that are in the debug files are MCCP codes. When posting text from the debug files, be sure to post text from the *.TXT file and *NOT* from the *.RAW file. Then you won't get these control codes embedded into your forum posts that we can't read.

SubAdmin Joined: 18 Nov 2001 Posts: 5187

I was able to reproduce a crash at that location with the same read address value. The call stack goes through many of the same functions right before the crash. It probably is the same bug, but a very different way of finding it.