<br><font size=2 face="sans-serif">Benjamin Herrenschmidt wrote:</font>
<br><font size=2><tt>> On Fri, 2006-10-20 at 16:04 +0200, Ulrich Weigand
wrote:<br>
> > <br>
> > Dwayne McConnell wrote:<br>
> > <br>
> > > I think %0xllx is the way to go. I would even advocate changing
<br>
> > > signal1_type and signal2_type unless it is actually too
dangerous.<br>
> > Is <br>
> > > there even a case where changing from %llu to %0xllx would
break<br>
> > things? <br>
> > > Perhaps with the combination of a old library with a new
kernel? <br>
> > <br>
> > Yes, these files are in fact used, and a change would break things.
<br>
> <br>
> In the case of output in hex without 0x, well, those files are used
by<br>
> code that is either assuming it's hex or is already broken. Adding
0x<br>
> might not actually break things, it depends what the use code does.
If<br>
> such code does strto(u)l, then adding 0x will still work.<br>
</tt></font>
<br><font size=2><tt>The two files I was commenting about currently use
%llu, i.e. they</tt></font>
<br><font size=2><tt>output in *decimal*. Changing this to hex may
break user space.</tt></font>
<br>
<br><font size=2><tt>This is a different situation from the *other* files,
which were using</tt></font>
<br><font size=2><tt>%llx, i.e. output in hex without 0x.</tt></font>
<br>
<br><font size=2 face="sans-serif"><br>
Mit freundlichen Gruessen / Best Regards<br>
<br>
Ulrich Weigand<br>
<br>
-- <br>
Dr. Ulrich Weigand<br>
GNU compiler/toolchain for Linux on System z and Cell BE<br>
IBM Deutschland Entwicklung GmbH, Schoenaicher Str. 220, 71032 Boeblingen<br>
Phone: +49-7031/16-3727 --- Email: Ulrich.Weigand@de.ibm.com</font><font size=2><tt><br>
</tt></font>