Bug 229903 – device-mapper init script fails for no-multilib 64-bit systems

• Product: Gentoo Linux
• Status: RESOLVED FIXED

[Description Albin Stjerna 2008-06-28 14:37:33 UTC]

The device-mapper init script is looking for librc.so in /lib/, while mine is in /lib64 since I'm using the no-multilib profile on amd64.

Reproducible: Always

Steps to Reproduce:
1. 2. 3.

[Comment 1 SpanKY 2008-10-26 02:55:12 UTC]

fixed in cvs

http://sources.gentoo.org/sys-fs/device-mapper/files/device-mapper.rc-1.02.22-r3?r1=1.5&r2=1.6

[Comment 2 Walter Meinl 2008-10-28 07:10:42 UTC]

(In reply to comment #1)
> fixed in cvs
> 
> http://sources.gentoo.org/sys-fs/device-mapper/files/device-mapper.rc-1.02.22-r3?r1=1.5&r2=1.6
> 
shouldn't that be in line 16 (start section)
 if [ ! -e /etc/init.d/root ] ; then
instead of
if [ -e /etc/init.d/root ] ; then

Today device-mapper-1.02.27 refused to start on my baselayout-2 system complaining I have baselayout-1

[Comment 3 Doug Goldstein (RETIRED) 2008-10-28 16:28:01 UTC]

(In reply to comment #2)
> (In reply to comment #1)
> > fixed in cvs
> > 
> > http://sources.gentoo.org/sys-fs/device-mapper/files/device-mapper.rc-1.02.22-r3?r1=1.5&r2=1.6
> > 
> shouldn't that be in line 16 (start section)
>  if [ ! -e /etc/init.d/root ] ; then
> instead of
> if [ -e /etc/init.d/root ] ; then
> 
> Today device-mapper-1.02.27 refused to start on my baselayout-2 system
> complaining I have baselayout-1
> 

You are correct. I've committed the change.

Additionally, I don't understand why this was a multilib issue since Gentoo requires that /lib be a symlink to /lib64 if that's what you're using and the -e check in bash will follow the symlink. On my system the actual library is in /lib64 and the check has worked fine for months.

[Comment 4 Seemant Kulleen 2008-12-18 21:40:48 UTC]

Hi Doug, I think the logic is a little weird in the scripts. Since this package is not designed to work with baselayout-1 at all, the depend() function should enable those depends when it detects baselayout-2.  The way it is setup in the current commit, it omits depends for baselayout-2 but adds them in for -1.

See what I'm saying? From the commit log, it looks like spanky reversed the logic in that function, but there's no indication as to why.

plzfixkthxbye

[Comment 5 Doug Goldstein (RETIRED) 2008-12-18 21:52:23 UTC]

Thanks for pointing that out Seemant. I've committed the fix.