Re: Setting up eth-multiplexer in QEMU

[Sergiu Ivanov]

Hello,

On Mon, Jun 29, 2009 at 10:18:35PM +0200, Thomas Schwinge wrote:
> On Mon, Jun 29, 2009 at 10:21:22PM +0300, Sergiu Ivanov wrote:
> > I've found out a stunning thing (for me) today: I checked out a fresh
> > Hurd from CVS in the following way:
> > 
> >  $ cvs -z3 -d:pserver:anonymous@cvs.savannah.gnu.org:/sources/hurd co hurd
> > 
> > Then I created build/ , went there, configured and ran make
> > pfinet. The executable file pfinet I got as a result behaves in the
> > *same* way as the modified pfinet + devnode! That is, things work a
> > little bit slowly, and wget stalls having downloaded the same 2,576
> > bytes :-(
> 
> May I ask, and I should have realized this earlier! -- in the context of
> our IRC discussion one or two weeks ago when we told you to update your
> system -- which version of GCC did you use to build pfinet?
> <http://savannah.gnu.org/bugs/?23551> -- did you use GCC 4.3?  If yes,
> then sorry for the wasted time!, and hooray for having (hopefully)
> resolved this mystery.

Yeah, hooray! :-) Downgrading to gcc-4.2 solved all my problems,
including the not-working modified pfinet.
 
> > Could somebody point out where I could find the correct
> > git repository?
> 
> <http://www.gnu.org/software/hurd/rules/source_repositories.html>: ``Git
> repositories on Savannah, see <http://git.savannah.gnu.org/cgit/> (search
> for hurd).'' -- <http://git.savannah.gnu.org/cgit/hurd/> is a more
> suitable link, which I'll add to that page.

Aha, great! Thank you!
 
> > Also, I didn't apply Debian patches, but I'm not sure whether they
> > could be the cause... I didn't apply them because, apparently,
> > something is different in the main Hurd branch and Zheng's branch,
> 
> Probably it's been some time since the HEAD branch has last been merged
> into Zheng's branch.  (This will improve as soon as I've transitioned
> these working-branches to Git.)

I see.
 
> > because exactly the series of patches which worked in Zheng's branch
> > failed in the source tree I got by means of the command mentioned
> > above.
> 
> Relevant (depending on your version of glibc) are (very likely) need the
> in6_addr.patch, to avoid a build failure, which is in HEAD (and Git
> master, but not in Zheng's branch, I guess).  Then there is the DHCP
> patch for pfinet, which is not important for your usage, and that's it:
> nothing else is relevant for pfinet.
 
That's clear. I'm using Zheng's branch again to build the tools he
created, so no problems as far as patches are concerned. Thank you for
the information!

Regards,
scolobb