Re: [PATCH] 2/6 Detect overflows in loop counters

Mingming Cao wrote:
> On Fri, 2006-08-18 at 12:45 -0500, Eric Sandeen wrote:
>> Mingming Cao wrote:

>> How much do we want to intermix full 32-bit fixes with the 64-bit 
>> changes?  All at once or step by step?  Step by step seems like a safer 
>> approach, though it probably means reworking some of the 64-bit patches 
>> (I'll sign up to do that work if 32-bit stuff goes in first...)
>>
> 
> Either way is fine with me ... 
> 
> But I think making the loop counter to unsigned value seems more
> straightforward than keep it as int type, and adding a check of < 0 in
> the end condition.

Well, the loop counters already -are- unsigned in these cases, via a 
typedef usually.  The issue is that if the loop tests that the counter 
is "<= <value>", where <value> is using all the bits in the type (say 
2^32-1 blocks....) then the test can never be false, and the loop will 
run forever w/o the extra "&& counter > <something>" to detect a wraparound.

-Eric

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642