Re: [v4l-dvb-maintainer] [PATCH 3/3] V4L: cinergyT2, remove bad usage of ERESTARTSYS

[Randy Dunlap <randy.dunlap@oracle.com>]

On Wed, 10 Oct 2007 00:18:28 -0400 Michael Krufky wrote:

> Mauro Carvalho Chehab wrote:
> > Hi Jiri,
> >
> > Em Seg, 2007-10-08 às 13:41 +0100, Jiri Slaby escreveu:
> >   
> >> cinergyT2, remove bad usage of ERESTARTSYS
> >>
> >> test of cinergyt2->disconnect_pending doesn't ensure pending signal and so
> >> ERESTARTSYS would reach userspace, which is not permitted. Change it to
> >> EAGAIN
> >>
> >>     
> >
> > checkpatch.pl is complaining about your changeset:
> >
> > do not use assignment in if condition
> > #82: FILE: drivers/media/dvb/cinergyT2/cinergyT2.c:492:
> > +     if ((err = mutex_lock_interruptible(&cinergyt2->wq_sem)))
> >
> > do not use assignment in if condition
> > #86: FILE: drivers/media/dvb/cinergyT2/cinergyT2.c:495:
> > +     if ((err = mutex_lock_interruptible(&cinergyt2->sem)))
> >
> > do not use assignment in if condition
> > #133: FILE: drivers/media/dvb/cinergyT2/cinergyT2.c:1036:
> > +     if ((err = mutex_lock_interruptible(&cinergyt2->wq_sem)))
> >
> > do not use assignment in if condition
> > #137: FILE: drivers/media/dvb/cinergyT2/cinergyT2.c:1039:
> > +     if ((err = mutex_lock_interruptible(&cinergyt2->sem)))
> 
> Is this illegal as per kernel codingstyle?  I could understand if we may
> want to avoid this sort of thing for the sake of code readability, but
> this seems 100% proper to me, especially considering that we're simply
> trying to catch an error return code.

Anyway, it's not strictly from CodingStyle.  The closest that
CodingStyle has to say about it IMO is this:

"Don't put multiple assignments on a single line either.  Kernel coding style
is super simple.  Avoid tricky expressions."

Also, Andrew Morton and Christoph Hellwig push for splitting up
the if-assignment lines, so it's a trend over the past few
(probably) years.  It's not real new.


> One of the things that I really enjoy about the c programming language
> is the fact that you can string many operations together into a single
> statement through the use of logic.  I hate the thought of a patch being
> nacked because of the above.  :-/

so you like the challenge of reading obfuscated code?

At any rate, it's rare that a patch is nacked only due to coding style,
unless it's blatant and occurs many times (like throughout the entire
source file).


> If this is indeed kernel codingstyle, then IMHO, we should let it slide
> for catching error return codes.

Nope.

---
~Randy