Re: remote debugging stubs (neighbourhurds)

[Greg Buchholz]

On Tue, 9 Sep 2003, Marcus Brinkmann wrote:

> On Mon, Sep 08, 2003 at 11:10:35PM +0200, Marco Gerards wrote:
> >
> > This is weird. Is /hurd/exec on the fs. Is there a /server/* ?  Attach
> > gdb to the fs. to see what it its loading.

        Yes, there is a /hurd/exec, and I have a /server directory (and
there is a 0 byte exec file in it).  In fact I dumped most of my hurd
partition to CD just to see what would happen if I ever got it to boot.

>
> Well, one thing that happens in the boot procedure is that exec is attached
> to the /servers/exec node.  This might only work on filesystems with write
> support.  In other words: there might be gaps in iso9660fs's support for the
> bootstrap procedure.

        Just for reference, when iso9660fs boots for real, it starts to
barf in libstore/open.c after it gets called from store_parsed_open().
Or it probably barfs here -- this is where I got tired of the "debug by
printf" method and started looking towards gdb and "boot".  If true, that
probably narrows it down to either file_name_lookup() or store_create()
that fails.


Thanks,

Greg Buchholz