[PATCH 2/2] staging: dgnc: switch timeout to signed type

Nicholas Mc Guire der.herr at hofr.at
Fri May 29 19:13:37 UTC 2015


On Fri, 29 May 2015, Dan Carpenter wrote:

> On Fri, May 29, 2015 at 07:21:26PM +0200, Nicholas Mc Guire wrote:
> > On Fri, 29 May 2015, Dan Carpenter wrote:
> > 
> > > On Fri, May 29, 2015 at 06:41:28PM +0200, Nicholas Mc Guire wrote:
> > > > The schedule_timeout*() helpers take the timeout as signed long, as
> > > > ch_close_delay in struct channel_t was not used for other purposes its
> > > > type was switched to signed long and the declarations fixed up.
> > > 
> > > Uh, we never pass it to schedule_timeout etc and even if we did how
> > > would that matter?  It's either 250 or 0.
> > > 
> > > What is the bug you are trying to fix and we can help you?
> > >
> > static code checkers being unhappy with type mismatch
> > automatic type conversion is ok if necessary but in this
> > case it simply is not as the ch_close_delay is only being
> > used in this one place so why not do it type clean ?
> 
> This seems like a pointless warning.  What does the warning look like?
> We pass ms to msecs_to_jiffies() and not to schedule_timeout() so it
> seems like somewhere something is confused.

Not really - just my carelessness - the msecs_to_jiffies was not in there
and I fixed up the types first - then put the msecs_to_jiffies in there
to fix up the time conversion ...oh well took the type conversion out
just to put it back in my self...sorry thats a bit braindead.
thanks for catching that.

> 
> > I'll turn the question around - what reason would there be to
> > go through type conversion if it is not needed ?
> 
> You can go crazy if you do ever pointless change which a static analysis
> tool suggests...
> 
> Btw, Smatch says that "ms" is always 250 here, actually.  I was guessing
> earlier when I said it could be zero.  Get a smarter static checker
> which can read code. 

wont blame it on coccinelle - its my scripts that are to blame - but in 
this case it was the cleanup after the fix for the warning that broke
it.

so 2/2 is pointless - sorry for that - pleas just toss it.

thx!
hofrat


More information about the devel mailing list