On Thu, Apr 04, 2019 at 12:57:29PM +0100, Ben Dooks wrote: > I've updated the series with the last round of review comments. > I think this is close to mergable now. Thanks. > Notes: > [1] we could do with some sort of -Wformat-printk for kernel printk ext > [2] sscanf and friends should be added at somepoint > [3] but these shouldn't be stopping an initial merge I agree. > for [1], should we consider a __attribute__(linux-printk) as well as > the formatting patch? Another attribute to specifiy the the __attribute__((format(printf,m,a))) is not only checking pure printf format but also the kernel's extensions? The yes, maybe but it may be implicit. > the formatting patch? it seems that today's kernel has a lot of extras > or should we look at some form of attribute allowing extensions to be > added, like __attribute__(format-extension(%pF,struct mac_address *) Yes, something like that will be needed and the mere presence of it would imply the __attribute__(linux-printk). But I don't think the attribute syntax is really appropriate (even if it would be very flexible). I was thinking about something like a few global #pragma sparse_printf_format "%pF" struct mac_address * but: 1) pragmas are ugly/people dislike pragmas 2) there may be (more) problems with the type declaration -- Luc
