On Mon, 2025-03-17 at 20:50 +0100, Miguel Ojeda wrote:
> On Mon, Mar 17, 2025 at 6:52 AM Joe Perches <joe@xxxxxxxxxxx> wrote:
> >
> > Not sure it's a 'real' need/issue as any new KERN_<LEVEL> that's
> > not KERN_CONT will already add a new line.
>
> But it is still intended that one uses an explicit newline, no?
That's my preference, yes.
> I am
> not sure I follow your other comment, so I may be missing something
> though.
My understanding is that
printk(KERN_INFO "foo"); // or pr_info("foo");
//with some long time before
printk(KERN_INFO "bar"); // or pr_info("bar");
will defer emitting "foo" and will append a newline to "foo"
when emitted only after the printk for "bar" is performed.
So it's generally useful to always use a newline for
complete messages as those are never deferred until
another printk is done.
> Regarding complexity, agreed, it is fairly involved -- I think the
> continuation case could be ignored, since it is rare and it is just a
> warning anyway.
My belief as well.
There are a number of macros and functions that are printk equivalents
that append newlines automatically and as well some pr_fmt uses
that include newlines too so it's almost impossible to get checkpatch
to work correctly anyway.
Things like:
fs/bcachefs/bcachefs.h:#define pr_fmt(fmt) "bcachefs: %s() " fmt "\n", __func__
