Re: [PATCH] daemon.c: mark a file-local symbol as static

From: Jeff King <peff@peff.net>
Date: Feb 22, 2016, 21:18 UTC
Message-ID: <20160222211827.GB15595@sigill.intra.peff.net>
In-Reply-To: <56CA5753.9030308@ramsayjones.plus.com>

On Mon, Feb 22, 2016 at 12:33:23AM +0000, Ramsay Jones wrote:

Show 12 quoted lines

> > Thanks, will do. You notice these with sparse, as I recall? I've meant
> > to look into running that myself, but it looks like we are not
> > warning-free with sparse currently. I see complaints like:
> > 
> >   connect.c:377:40: warning: incorrect type in argument 2 (invalid types)
> >   connect.c:377:40:    expected union __CONST_SOCKADDR_ARG [usertype] __addr
> >   connect.c:377:40:    got struct sockaddr *ai_addr
> > 
> > As far as I can tell, that's just noise. Do you have a ready-made recipe
> > for silencing it?
> 
> Ah, I think you must be on a very old version of sparse.

I have whatever comes with debian unstable. Looks like 0.5.0 from November. So not _that_ old, but I can well believe it is missing tweaks found at the tip of their development.

I guess I was wondering whether I should be adding "make sparse" to my set of pre-submission checks. But I can't say I'm enthused about manually keeping another tool up to date. :)

> On Linux, there is a single warning, which results from a hard-coded
> value (max size of memcpy et. al., used for the kernel) which should
> at least be settable from the command line (I have a patch somewhere
> which I have not sent upstream).

Yeah, I saw that one, for a large zeroing memset of a heap buffer. I don't think it's _wrong_, though it would probably not hurt to use xcalloc instead.

Thanks for the pointers.

-Peff