Re: [Gumstix-users] flushing stdout and kill/interrupt

[Craig H. <cr...@gu...>]

On May 23, 2005, at 10:11 PM, Darren Gibbs wrote:

>
> On May 23, 2005, at 6:39 PM, Dave Hylands wrote:
>
>>> Just to be entertaining, it seems like it's not as black and  
>>> white as
>>> I at first thought... it's working sometimes... 8/10 times  
>>> everything
>>> is fine, 2/10 times ttyS0 isn't taking input, but everything else is
>>> fine.  So far this only happens when the ports are configured and  
>>> the
>>> app is launched by an init script.  Launching the app by hand
>>> (without changing the port configurations) seems to work everytime.
>>>
>>>
>>
>> Yep - that's definitely weird. The other difference between running
>> from init.d and when running from a shell, is who your parent process
>> is, and where stdin/stdout is directed.
>>
>> I think that programs launched from init will have stdin and stdout
>> set to /dev/console.
>>
>
> Now that I have the signal handler in place I used the launch  
> script to redirect my app's stdout/stderr to a log file in /tmp.   
> If things aren't working I can log in and kill the app with 'kill  
> -1" and then read the log.  Remarkably, now that I have the debug  
> mechanism in place, I can't get the app to fail again!

I had a similar experience trying to figure out why sdpd iirc (part  
of the bluetooth utils) wasn't running properly from the startup  
script, but would run fine from command line.  I don't think I did  
ever figure out what the problem was, but adding a sleep at the  
appropriate time I think cured it.

C