[asterisk-users] zaptel 1.4.10 regression with TE220B on Proliant DL380 G5 ?

[Matthew Fredrickson]

Ex Vito wrote:
> On Wed, Apr 16, 2008 at 7:18 PM, Matthew Fredrickson <creslin at digium.com> wrote:
>> Ex Vito wrote:
>>>  Tested with no 4K stack kernel and stackcleanup svn branch
>>>  zaptel version. Correct, the kernel no longer "complains" about
>>>  the soft hangup.
>>>
>>>  However the system still hangs (console inoperative, etc) while
>>>  ztcfg'ing...
>>>
>>  That is normal while the firmware is loading.  It should go away after the
>>  firmware has loaded.
>>
> 
>   Ok. So here is our reasoning according to collected info. Please
>   correct us where appropriate:
> 
>   1. The system is supposed to hang while the firmware loads into
>       the DSPs under any zaptel version
>   2. zaptel 1.4.10 leads to a soft hangup detected, zaptel 1.4.9.2
>      does not (assuming softhangup detection active in kernel)
>   3. zaptel 1.4.10 takes much longer ztcfg'ing than 1.4.9.2, that's
>       why the soft hangup is detected under zaptel 1.4.10
>       (difficult to time, but let's say 1.4.10 takes 10s, 1.4.9.2
>        takes 3s)
> 
>   Now, back to the original question:
> 
>   - Should this be considered a regression ?
>   - Next steps:
>     a) file a bug and move this analysis to the bug tracker
>     b) don't file bug and move analysis to the dev list
>     c) don't file bug, keep on working on the users list
> 
>>  I recommend 1.4.10 by default.  However, from what you said it would appear
>>  that you are having problems with 1.4.10 so you might stay with 1.4.10 if
>>  you are not having any issues with it.

I just realized where this is coming from.  I was attempting to patch 
this from a different angle, but as soon as you mentioned the drastic 
difference in load time I realized what had happened.  I'm going to make 
another update to my stack reduction branch to see if I can fix this. 
I'll let you know when it's done.

-- 
Matthew Fredrickson
Software/Firmware Engineer
Digium, Inc.