Re: [PATCH/RFC] Documentation: git-daemon inetd configuration fix

[Matthias Kestenholz <matthias@spinlock.ch>]

On Fri, 2007-03-16 at 01:43 -0700, Junio C Hamano wrote:
> Matthias Kestenholz <matthias@spinlock.ch> writes:
> 
> >  On my debian server, this change was needed to get git-daemon running.
> >  I am no inetd expert, so I don't know where the error was.
> >
> >  Perhaps this patch might help someone else.
> >
> >
> > diff --git a/Documentation/git-daemon.txt b/Documentation/git-daemon.txt
> > index 9ddab71..499c1f3 100644
> > --- a/Documentation/git-daemon.txt
> > +++ b/Documentation/git-daemon.txt
> > @@ -177,7 +177,7 @@ git-daemon as inetd server::
> >  +
> >  ------------------------------------------------
> >  	git stream tcp nowait nobody  /usr/bin/git-daemon
> > -		git-daemon --inetd --verbose --export-all
> > +		/usr/bin/git-daemon --inetd --verbose --export-all
> >  		/pub/foo /pub/bar
> >  ------------------------------------------------
> 
> That sounds _very_ wrong.  What you changed is argv[0] (i.e. the
> beginning of "service program arguments" field) and it should
> not matter what you have there for inetd to launch git-daemon.
> Presumably your inetd runs with /usr/bin in your path so even
> the "service program" field (the one after 'nobody') should not
> need the full path (but it is a good idea to be explicit in that
> field as a principle), but needing the full path in argv[0]?
> 
> On _my_ debian box, I do not need a full path there, so it is
> not even Debian in general but problem is on your particular
> box.  Can you dig a bit more to see what you are doing any
> differently?
> 
> For example, do you really mean the "service program arguments"
> field that matters?
> 


Uhm... I don't get it. I changed the second /usr/bin/git-daemon back to
git-dameon (and even to "foo") and it continues working after restarting
inetd.

I cannot reproduce the problem I had earlier, but everything works fine
with the line from the docs now.

Btw, git is installed in /usr/bin , so that cannot be the reason for the
failure I saw.


Oh well, sorry for the noise and thanks for the fast response.
	Matthias