[sheepdog] [PATCH] lib: choose a correct output in logger

[Liu Yuan]

On Wed, Jul 17, 2013 at 04:48:10PM +0900, Hitoshi Mitake wrote:
> At Tue, 16 Jul 2013 15:01:07 +0800,
> Kai Zhang wrote:
> > 
> > 
> > On Jul 15, 2013, at 9:03 PM, Hitoshi Mitake <mitake.hitoshi at gmail.com> wrote:
> > 
> > > Current logger chooses stderr as an output of logs in cases
> > > logarea_init() isn't called or default formatter is used.
> > > 
> > > This patch lets logger choose stderr when a priority of logs is less
> > > than or equal to LOG_NOTICE. If the priority is larger than or equal
> > > to SDOG_INFO, stdout is used.
> > > 
> > > Signed-off-by: Hitoshi Mitake <mitake.hitoshi at lab.ntt.co.jp>
> > > ---
> > > lib/logger.c |    6 ++++--
> > > 1 files changed, 4 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/lib/logger.c b/lib/logger.c
> > > index 253163d..059140f 100644
> > > --- a/lib/logger.c
> > > +++ b/lib/logger.c
> > > @@ -407,8 +407,10 @@ static notrace void dolog(int prio, const char *func, int line,
> > > 		memset(msg, 0, sizeof(struct logmsg));
> > > 		init_logmsg(msg, &tv, prio, func, line);
> > > 		len = format->formatter(str_final, MAX_MSG_SIZE, msg);
> > > -		xwrite(fileno(stderr), str_final, len);
> > > -		fflush(stderr);
> > > +
> > > +		FILE *f = SDOG_INFO <= prio ? stdout : stderr;
> > > +		xwrite(fileno(f), str_final, len);
> > > +		fflush(f);
> > > 	}
> > > }
> > > 
> > 
> > It is odd to declare a variable 'f' in the middle of the method.
> 
> No. We employed gnu99 style in 7141b4535bad, interleaving declarations
> and statements are allowed.
> 
> Declaring variables at top of functions has no meaning. Narrowed
> scope of variables improves readability significantly.
> 

What do you mean by 'improves readablility significantly'? It seems that you
assume we write big complex function, in which case declaration in the middle
might help people get the idea what the variable is. But this assumption is
fundamentally wrong to me. Trust me, big function will always be harder to
understand than partitioning it into smaller well organized functions. Any
effort of trying to improve the readability of a big & complex function will be
in vain.

Thanks
Yuan