Hi Guenter, On Fri, 19 Aug 2011 07:16:39 -0700, Guenter Roeck wrote: > On Fri, Aug 19, 2011 at 03:07:56AM -0400, Jean Delvare wrote: > > I am under the impression that error paths in functions > > aem_init_aem1_inst() and aem_init_aem2_inst() are incorrect. In > > several cases, the function returns 0 on error, which I suspect is > > not intended. Fix this by properly tracking error codes. > > > > Signed-off-by: Jean Delvare <khali@xxxxxxxxxxxx> > > Cc: Darrick J. Wong <djwong@xxxxxxxxxx> > > Cc: Guenter Roeck <guenter.roeck@xxxxxxxxxxxx> > > --- > > Darrick, can you please confirm that none of the missing error codes > > was on purpose, and that my patch doesn't break anything? Thanks. > > with your new code, if instance <n> fails, subsequent instances will not be initialized. > Already initialized instances will not be removed. So you end up with instances > 0 .. <n-1>. In the old code, it was possible to end up with instances 0 .. <n-1> <n+1> ... <m>. > So there is definitely a behavioral change. Thanks for the review. The behavioral change is pretty much intended, as my assumption is that the original code didn't do the right thing. > On the other side, it is odd that some errors would cause instantiation to abort, while > instantiation would continue for other errors. Agreed. In fact I think it is wrong that the failure to initialize one instance has any impact on other instances, no matter what error happened. That would be a separate bug though, which my changes are simply making more visible. I'll send a separate patch for it. -- Jean Delvare _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
