Re: [RFC PATCH v4 03/26] pkt-line: Make packet_read_line easier to debug

From: David Brown <davidb@quicinc.com>
Date: Oct 30, 2009, 19:31 UTC
Message-ID: <20091030193108.GA30397@huya.quicinc.com>
In-Reply-To: <20091030191239.GF10505@spearce.org>

On Fri, Oct 30, 2009 at 12:12:40PM -0700, Shawn O. Pearce wrote:

Show 8 quoted lines

> > If you really just want to initialize to zero, using
> > 
> >   char linelen[5] = { 0 };
> 
> Bleh, I find that has hard to grok as what we have now.  Perhaps my
> understanding of the relevant standards is incomplete, but I'd read
> that as linelen[0] = 0, but the other 4 positions are undefined
> and may be not be initialized.

C isn't that into undefined things. In this particular case, all remaining values will be initialized to zero.

Show 6 quoted lines

> I actually considered this one, but again I wasn't clear what would
> happen in the standard C library when we fed a string that wasn't
> actually NUL terminated.  Is the library permitted to call strlen()
> before formatting?  If so strlen() could SIGSEGV if we are unlucky
> and no NUL is present between our string and the end of the assigned
> memory region.

The linux manpage says "If a precision is given, no null byte need be present". This text is copied verbatim out of the Posix specification, so a C library that failed to handle this would be non-compliant.

I think the %.4s is clearest, and is, in fact, somewhat idiomatic.

David