> > OK... Here we go. I've converted all drivers in our CVS repository > > to a new model where we do the two kmallocs separately. I didn't > > commit it yet because it's a very important change (47 chip drivers > > affected). So I'd appreciate code reviews and tests before I apply > > it. > > > > The patch is here: > > http://jdelvare.net1.nerim.net/sensors/lm_sensors-CVS-separate-mallocs.diff > > It applies on latest CVS. > > > > I tested only the eeprom and lm90 drivers so far. Seems to work, but > > the changes are error-prone so what it "seems" is hardly enough to > > be sure it's correct. > > At a quick glance it looks sane. Well, I doubt that a quick glance is enough to ensure that the patch didn't break anything. I'm confident that the memory allocation and "normal" freeing are correct. What I consider error-prone is the error handling in detect functions. In several cases I had to rename labels. It's hard to be sure that all error paths are absolutely correct, although I did my best not to screw anything up. -- Jean Delvare http://www.ensicaen.ismra.fr/~delvare/
