Hi Daniel, On Wed, Aug 24, 2016 at 12:21:23PM +1000, Daniel Axtens wrote: > Hi Simon, > > > The ckpt_regs usage in gpr32_set_common/gpr32_get_common() > > will lead to cppcheck error. > > > > [arch/powerpc/kernel/ptrace.c:2062]: (error) Uninitialized variable: > > ckpt_regs > > [arch/powerpc/kernel/ptrace.c:2130]: (error) Uninitialized variable: > > ckpt_regs > > > > A straightforward fix to clean it. > > I'm always happy to see cppcheck warnings fixed :) Thanks for raising this issue :)
> > > static int gpr32_get_common(struct task_struct *target, > > const struct user_regset *regset, > > unsigned int pos, unsigned int count, > > +#ifdef CONFIG_PPC_TRANSACTIONAL_MEM > > void *kbuf, void __user *ubuf, bool tm_active) > > +#else > > + void *kbuf, void __user *ubuf) > > +#endif > > I wonder if it might be possible to avoid some of the ifdefs and general > churn by making the tm_active argument __maybe_unused rather than > ifdefing around it? > > In particular, it would mean the two hunks in the function definitions > and these these two hunks at the call site would be unnecessary: I think keeping tm_active argument for "ifndef CONFIG_PPC_TRANSACTIONAL_MEM" case (with __maybe_unused prefix) will be somehow strange -- Whatever value is provided in the caller function for tm_active, programmer might be puzzled and cost sometime to think about it. I don't like to use "__maybe_unused" to bypass this warning. Thanks, - Simon
