[debian-devel:13510] FWD: [Re: Arm bf problems - termwrap and null strings]

[Taketoshi Sano <kgh12351@xxxxxxxxxxx>]

佐野＠浜松です。

 debian-boot@lists.debian.org から転載。

 termwrap の問題は Adam が解決してくれたようです。

######################################################################
##      Set some environment variables.
######################################################################
# reads /etc/environment.
test -f /etc/environment && . /etc/environment

ですね。

In <oabstin2z7.fsf@xxxxxxxxxx>,
  on "07 Jan 2001 22:21:16 -0500',
   with "Re: Arm bf problems - termwrap and null strings",
 Adam Di Carlo <adam@xxxxxxxxxxx> さん wrote:

> Wookey <wookey@xxxxxxxxxxxx> writes:
> 
> > I've got some Arm boot-floppies (from CVS about 10 days ago ~ 2.2.19) which
> > nearly work. I've fixed a couple of problems and I'll post diffs here when I
> > get a set that work.
> 
> Sure.  Do you have (and, if not, want) direct CVS access?
> 
> > Current problems are:
> > 1) inittab.install in root/bin is the one that gets put in /etc for the
> > second-stage install. This runs dpkg-reconfigure with /sbin/termwrap. This
> > doesn't work on arm. the problem seems to be that termwrap needs a file
> > called /etc/environment, and that isn't present in root/bin or base.tgz. I've
> > worked around it by simply removing this from the inittab,
> 
> I've got a better workaround -- I've changed the script to only source
> /etc/environment if the file actually exists.  Done in CVS.
> 
> > 2) dbootstrap falls over trying to set the hostname. If you kill it (so it
> > restarts) then it continues but /etc/hostname ends up being "(null)" 9which
> > caues more problems later), so I suspect a similar problem to the one below
> > but haven't gone into the detail yet - that's next to investigate, straight
> > after this mail. I just wonder if this is a prob others found which is now
> > fixed in CVS?
> 
> Not that I know of.  I suggest we keep looking into this.
> 
> > Fixed problems are: in utilities/dbootstrap/bootconfig.c. dbootsrap bombs out
> > with a segfault when you select 'make drive bootable'. This is because
> > 'Arch2' is null for arm when it gets tested to see if it is 'acorn' or
> > 'netwinder' in order to decide which message explaining why this option is
> > pointless to print :-). (not sure why it segfaults rather than justs prints
> > neither message). This is fixed in utilities/dbootstrap/main.c in
> > (get_arm_model which needs 
> >     if (!strcmp(buf, "acorn-", 6)) {
> > changing to 
> >     if (!strncasecmp(buf, "acorn-", 6)) {
> > 
> > Also I note this prob always occurs if you use bootconfig_test as the
> > get_arm_model fn in main is not available. This must also
> > be true for get_alpha_model which was already in bootconfig.c for the
> > _TESTING_ case, presumably someone else's optimistic attempt to get it to
> > work :-)
> 
> Sounds good, looking forward to your patch.  Hope it comes soon.
> 
> -- 
> .....Adam Di Carlo....adam@xxxxxxxxxxxxxxxx<URL:http://www.onShore.com/>

-- 
     ＃ (わたしのおうちは浜松市、「夜のお菓子」で有名さ。)
    <kgh12351@xxxxxxxxxxx> : Taketoshi Sano (佐野　武俊)