Hi Jean, On 27 August 2013 19:43, Jean Delvare <khali@xxxxxxxxxxxx> wrote: > On Tue, 27 Aug 2013 15:10:43 +0100, Russell King - ARM Linux wrote: >> On Tue, Aug 27, 2013 at 03:57:53PM +0200, Jean Delvare wrote: >> > Also note that the fix Sachin sent for acpi_power_meter is slightly >> > different from what Russell complained about originally. Russell >> > complained about: >> > >> > struct __initdata foo bar; >> > >> > While acpi_power_meter uses: >> > >> > struct foo __initdata bar; >> > >> > The former does indeed NOT work in my tests, so Russell was right >> > complaining. The good news is that there are only a few dozen >> > occurrences of this in the kernel tree. >> >> Indeed - and these "don't work" ones are definitely worth fixing. > > Agreed. > >> > The later OTOH works fine in my tests so I see no reason to change it - >> > unless someone can find a gcc version and platform combination where it >> > does not work. Hope this clarifies the whole situation... >> >> Yes, better not to change what's already there if it works fine. >> However, as I mentioned in my previous email, it's probably better to >> have a convention here for new code rather than having people stick >> these markers in random places within the declaration and hit the "it >> silently doesn't work" case. > > I totally agree, I think we want a checkpatch rule for this. I'll ask > for it in a separate thread. I had already asked Joe to add this as a checkpatch rule. He mentioned the following (which I quote): "There's no way for checkpatch to look at an __<foo> use and determine it should be before or after a variable. __scanf, __printf, __cold, etc are often place before declarations." -- With warm regards, Sachin _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
