timestamp_before_wait might be used uninitialized in this function

[walter harms]

Denys Vlasenko schrieb:
> On Tuesday 13 January 2009 19:21, walter harms wrote:
>> Cristian Ionescu-Idbohrn schrieb:
>>> I see this warning, and I don't like it:
>>>
>>> busybox-1.13.2/networking/udhcp/dhcpc.c:
>>> In function `udhcpc_main':
>>> busybox-1.13.2/networking/udhcp/dhcpc.c:321:
>>> warning: `timestamp_before_wait' might be used uninitialized in this function
>>>
>>> Is this patch appropriate?
>>>
>>> --- busybox-1.13.2/networking/udhcp/dhcpc.c.~1~	2008-11-09 18:27:58.000000000 +0100
>>> +++ busybox-1.13.2/networking/udhcp/dhcpc.c	2009-01-13 13:09:30.000000000 +0100
>>> @@ -342,7 +342,8 @@
>>>  				/* Else: an error occured, panic! */
>>>  				bb_perror_msg_and_die("select");
>>>  			}
>>> -		}
>>> +		} else
>>> +			timestamp_before_wait = 0;
>>>
>>>  		/* If timeout dropped to zero, time to become active:
>>>  		 * resend discover/renew/whatever
>>>
>>>
>>> Cheers,
>>>
>> maybe a simple
>>   unsigned timestamp_before_wait=0;
>>
>> will do the same ?
> 
> Yes. but
> 
> unsigned timestamp_before_wait = timestamp_before_wait;
> 
> is better wrt code size. :)
> 
> fixed. thanks.
> 

aehm, that will only fool the compiler to thing it is used.
timestamp_before_wait=0 will make sure that it works even if the 'if (tv.tv_sec > 0)' fails.
and it will fail because:

        timeout = 0;
        already_waited_sec = 0;
...
        tv.tv_sec = timeout - already_waited_sec;		
	if (tv.tv_sec > 0)
		timestamp_before_wait = (unsigned)monotonic_sec();
...	
	already_waited_sec += (unsigned)monotonic_sec() - timestamp_before_wait;



>> btw: can unsigned be replaced with unsigned int or whatever is useful here ?
> 
> ?! "unsigned" and "unsigned int" are the same.

  this is true. a compiler has no problems but i prefer clear types.