[16237] in Athena Bugs
Re: Case 118859: sgi [8.2.9]: bringing some things local with mkserv
daemon@ATHENA.MIT.EDU (Robert A Basch)
Wed Aug 26 17:16:47 1998
To: bugs@MIT.EDU
Cc: rbasch@MIT.EDU, JWEISS@MIT.EDU
In-Reply-To: Your message of "Wed, 26 Aug 1998 10:28:27 EDT."
<199808261428.KAA11744@team-usa.MIT.EDU>
Date: Wed, 26 Aug 1998 17:13:36 EDT
From: Robert A Basch <rbasch@MIT.EDU>
> What should have happened?
> the rootlinks should have been updated to point to /var/server
> deliverator# ls -l {,/os,/var/server}/usr/bin/strip
> lrwxr-xr-x 1 root sys 17 Mar 18 19:32 /os/usr/bin/strip ->
> ../lib/driverwrap
> lrwxr-xr-x 1 root sys 17 Aug 22 02:28 /usr/bin/strip ->
> /os/usr/bin/strip
> lrwxrwxrwx 1 root sys 17 Aug 22 05:53
> /var/server/usr/bin/strip -> ../lib/driverwrap
>
> We see that the link is brought local as it should be, but
> /usr/bin/strip still points at the packs. Arguably this is a mkserv
> bug, but due to complications with relative and absolute links, etc.
> I'm not convinced it's possible to get mkserv to DTRT in all cases.
> In the past we've attempted to avoid cases like this by having track
> copy the link on the os local rather than linking it it. However,
> using the native SGI install stuff, I'm not sure that is practical
> either.
So, in your example, you're saying that the install should leave
/usr/bin/strip as a symlink to ../lib/driverwrap, rather than a link
to /os/usr/bin/strip? I have been thinking that this makes a certain
amount of sense, but it would require tweaking the install volume
distributions to make inst do this, lots of testing, etc. We might
consider this for 6.5.
> Anyone have any thoughts or suggestions?
Why can't mkserv copy the link, e.g. leaving /usr/bin/strip a link to
../lib/driverwrap, just as is /var/server/usr/bin/strip?
Bob
