Re: [PATCH 4/8] staging: et131x: Remove ununsed statistics

[Jeff King <peff@peff.net>]

On Sat, Sep 13, 2014 at 08:45:56AM -0700, Greg KH wrote:

> On Sat, Sep 13, 2014 at 12:37:46PM +0300, Dan Carpenter wrote:
> > On Thu, Sep 11, 2014 at 10:59:42PM +0100, Mark Einon wrote:
> > > >From struct ce_stats; unicast_pkts_rcvd, unicast_pkts_xmtd,
> > > multicast_pkts_xmtd, broadcast_pkts_rcvd and broadcast_pkts_xmtd
> > 
> > For some reason something adds a '>' to the start of lines which start
> > with 'From'.  I don't know what it is...
> 
> It's an email protocol requirement, some RFC dictates it as "From" at
> the start of the line is an email "start" flag.

It's not an RFC thing. It's a side effect of the mbox format, which
squashes together multiple messages with "From " lines to mark their
starts. So many mbox implementations will quote them as ">From" (others
introduce a Content-Length header, or are simply more careful about
making sure that the line looks like a real "From " line, which should
contain a date).

If somebody's MUA is actually transmitting emails with the quoting,
that's wrong. It is a local storage problem, and they should not be
spreading the quoting disease to other systems.

> > When I apply this patch with 'git am' then it just removes the From
> > line.
> 
> As it should :)

That seems wrong. We should either leave it as-is (i.e., assume the
writer used no quoting and really did mean ">From") or strip the ">" to
turn it into "From" (i.e., assume the writer did use quoting). In some
implementations, a literal ">From" gets quoted to ">>From" and so on. So
we could even strip one level of quoting from such things (if we assume
the writer was such an implementation).

I don't think we can make this 100% foolproof without knowing which mbox
variant the writer used. But dropping the line is probably the worst
possible thing, as it does not match _any_ variants. :)

-Peff