H120 crash I05: at310205F8

[Datman]

Saturday night H120 has crashed two times adjusting the gain during recording at 192kbps.

I05: at310205F8

Another time has crashed pressing STOP at the end of a recording (HD led was on after crashing).

Now I'll try the new software (2006 09 11). I'll let you know about it on next Monday.

[pabouk]

The problem is already reported in the bug tracker here:
#5961 crash while recording
Please add your new findigs there as a comment.

There are also related problems in the bug tracker:
#5952 crash while recording in wavpack
#5963 Crash when leaving "Recording Settings" entered from the FM Radio screen

These problems are related to the new codec recording architecture:
http://forums.rockbox.org/index.php?topic=6152.0

[Toni]

I don't think this is related to the new recording codec system. As I described in http://www.rockbox.org/tracker/task/5961 I could get perfectly working binaries after I changed to the configure script from before 3.Sept.06. I have no recent developement environment to test with  the old configure script. If someone has, it would be interesting, wether these problems persist with the combination of old configure and new compiler.

[Lear]

I05 means division by zero. That's one of the recording crashes, and not related to the codecs, but should be fairly easy to fix. I added a comment to your bug with some details about it.

[Toni]

OK, this one has been fixed:
  http://www.rockbox.org/tracker/task/5972
The patch prevents from DivideByZero in read_peak_levels().

[Alec]

When is this likely to be included in a daily?  The patch tracker still says
Due in Version: Undecided
Due Date: Undecided


Alternatively, what would I download to compile a build including this patch?

*Edit - I just realised I've got an old build - 060913. Maybe its already in.

[pabouk]

You can see that the task was closed successfully:

Closed by  Martin Scarratt (mmmm-)
Date:  Wednesday, 13 September 2006, 08:22PM
Reason for closing:  Fixed

On the related task with the patch solving the problem FS#5961 you can see that it was closed successfully too:

Closed by  Martin Scarratt (mmmm-)
Date:  Wednesday, 13 September 2006, 08:22PM
Reason for closing:  Fixed

and in the listing of CVS commits you can check that the patch was really committed:

13 Sep 18:21   Martin Scarratt   apps/recorder/recording.c 1.135     Fix a possible division by zero which caused much "IO5: at 310205A0" type angst while recording

Do you still have similar problems with recording in the current CVS build?

[Alec]

Thanks - I don't really know my way around a CVS.

I've only had a quick bash of recording with the latest build, but no problems so far.

[jhMikeS]

Yeah, AGC. Petur uses that. I wonder if it's something there. I'm running a test recording radio for a couple hours with AGC safety clip on. I'll see what I get.

Toni: Since the code is most familiar to both of us, perhaps you could spot some big time booboo I might have made. I'm not seeing anything obvious to me. I could be too close to it all though.

The crashes appear somewhat random where as other problems I diagnosed regarding memory handling in general before committing were very consistent in both their locations and the exception code (almost always an IllInstr in IRAM).

[jhMikeS]

That's a divide by zero error (DivX0) in RAM somewhere. Please write the desriptions too so I don't have to look them up   .

Just test recorded ~2h:17m and will check it out.

[jhMikeS]

Guess I should have noticed that Lear mentioned everything in September (!) and even fixed. Need to read more thoroughly at times and not respond when in a rush but when I've got time (like now). Go ahead a poke at me for it. I deserve it sometimes.  

[Mmmm]

  Nice one Mike!