Re: [patch] redboot networking

[Gary Thomas <gary at mlbassoc dot com>]

On Tue, 2006-03-14 at 16:41 +0100, Wolfgang=?iso-8859-1?Q?_K=F6bler?=
wrote:
> Hi,
> 
> I just notices I should have sent the attached email to this list instead of
> ecos-discuss.

Applied, thanks.

> 
> Bye,
> Wolfgang
> 
> email message attachment
> On Tue, 2006-03-14 at 16:41 +0100, Wolfgang=?iso-8859-1?Q?_K=F6bler?=
> wrote:
> > ceware.org>
> > List-Subscribe: <mailto:ecos-discuss-subscribe@ecos.sourceware.org>
> > List-Archive: <http://ecos.sourceware.org/ml/ecos-discuss/>
> > List-Post: <mailto:ecos-discuss@ecos.sourceware.org>
> > List-Help: <mailto:ecos-discuss-help@ecos.sourceware.org>,
> >  <http://ecos.sourceware.org/ml/#faqs>
> > Delivered-To: mailing list ecos-discuss@ecos.sourceware.org
> > XFMstatus: 0000
> > Sender: ecos-discuss-owner@ecos.sourceware.org
> > From: Wolfgang=?iso-8859-1?Q?_Köbler?= <wk-list@koebler.com>
> > To: ecos-discuss@ecos.sourceware.org
> > Subject: Re: [ECOS] [patch] redboot networking
> > 
> > Huh ? What killed the rest of my mail ? So second try:
> > 
> > Hi,
> > 
> > In ecos/packages/redboot/current/src/net/tcp.c there are the two functions
> > __tcp_write and __tcp_write_block . Both have the same interface and do the
> > same thing, but the second is the blocking version of the first. Both
> > return an integer. __tcp_write returns the number of bytes that are (going to
> > be) send. So __tcp_write_block could return the number of bytes that have been
> > sent. But it always returns 0 (or -1 as error code). In __tcp_write_block,
> > there is a variable called "total" that is initialized to zero and returned
> > (in case of no error). But this variable is never increased.
> > 
> > I think it is not meant to be this way. I think this variable (and hence the
> > return value) should contain the total number of bytes sent during this
> > function call. Changing this can be done with the patch below.
> > 
> > The patch below is a very small change. It probably increases the memory
> > footprint by a few single bytes. In current (2006-03-09) ecos CVS
> > references to __tcp_write_block can be found in:
> > /ecos/packages/redboot/current/include/net/net.h
> > /ecos/packages/redboot/current/src/net/http_client.c
> > /ecos/packages/redboot/current/src/net/net_io.c
> > /ecos/packages/redboot/current/src/net/tcp.c
> > (via "find -exec grep -il __tcp_write_block {} \;")
> > 
> > net.h
> >   should not be affected by this patch as it only includes the declaration
> > http_client.c
> >   should not be affected by this patch as it does not check the return 
> >   value.
> > net_io.c
> >   should not be affected by this patch. It includes code that uses
> >    __tcp_write_block as if the patch was already applied. But this has no
> >   effect (the assigned vars are overwritten or never used afterwards).
> > tcp.c
> >   is the file changed by the patch. It includes no additional references to
> >   __tcp_write_block, only the function definition.
> > 
> > 
> > Bye,
> > Wolfgang
> > 
> > 
> > --- ecos-orig-2006-03-09/ecos/packages/redboot/current/src/net/tcp.c   
> > 2004-09-24 14:51:22.000000000 +0200
> > +++ ecos/packages/redboot/current/src/net/tcp.c 2005-12-15 01:46:53.000000000
> > +0100
> > @@ -868,6 +868,7 @@
> >          if (n > 0) {
> >              len -= n;
> >              buf += n;
> > +            total += n;
> >          }
> >          __tcp_poll();
> >      }
> > 
> > 
> > 
> > --
> > Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
> > and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss
-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------