Re: printing aprint_error_count

[Masanobu SAITOH <msaitoh%execsw.org@localhost>]

On 2016/07/15 18:52, Martin Husemann wrote:
> On Fri, Jul 15, 2016 at 06:30:23PM +0900, Masanobu SAITOH wrote:
> >  Or enclose with #ifdef DIAGNOSTIC?
>
> Assuming we do not use aprint_error when there is no "error",

Some drivers use aprint_error() to warn a problem and continue
attaching.

> hiding it for
> non-DIAGNOSTIC kernels does not make sense to me (at least in the !silent
> or !quiet case).

Me neither though some aprint_error() might make user nervous.

> I wonder if we should make the error count available via a read-only sysctl,
> so userland could act on it later (especially in quiet or silent boots).

 It would be good, but if we change the return value of the
attach function from void to int(or others), it will become
less required.

 I'll commit without #ifdef DIAGNOSTIC.

> Martin


--
-----------------------------------------------
                SAITOH Masanobu (msaitoh%execsw.org@localhost
                                 msaitoh%netbsd.org@localhost)