Norman,
Hadn't thought about that; yes the manifest should contain paths with
the final installation location the user has chosen. For the binary RPMS
the installation location will be fixed but it could change for the
source RPMS as they can be relocated.
I will have to think about it some more but initially I think it is OK
if the prefix path gets put in the manifest always, and DESTDIR is
ignored. I in this case it was the mix of the prefix and DESTDIR that
caused the problem.
DESTDIR I assume just effectively redirects the installation to another
location but any processing commands used in the installation will still
use prefix, i.e. commands that change paths in scripts etc.. This means
that entries in the manifest should always get prefix paths.
Steve.
-----Original Message-----
From: Starlink development [mailto:[log in to unmask]] On Behalf Of
Norman Gray
Sent: 26 April 2005 11:43
To: [log in to unmask]
Subject: Re: Starlink Nightly Build On System:RHEL-WS3-3_i386
Peter and Steve,
On 2005 Apr 25 , at 19.40, Norman Gray wrote:
> I've just taken a look at those files and realised that the handling
of
> DESTDIR in the manifest lines doesn't seem to be entirely consistent,
> with it being omitted from some lines.
I've fixed the remaining omission, so that now the manifest should
reflect the installation location including any DESTDIR specified on
the make command line.
Doing this, it occurred to me that there is a case for consistently
_omitting_ DESTDIR from the manifest lines, on the grounds that the
manifest should reflect the final installation location, and not a --
presumably temporary -- staging location. In either case, the calling
process knows what the DESTDIR is and so can add or delete it as
necessary. It probably doesn't make a lot of difference either way, as
long as any final manifest has the correct locations, but the current
behaviour (including DESTDIR in the manifest) is probably a little less
surprising, and therefore preferable.
However this is easy to change now, and hard to change latter. So
Steve: since this principally affects you right now, can you comment on
which you'd find more reasonable, or would prefer. Or does anyone else
with rpm/.deb experience have an opinion here?
Norman
--
----------------------------------------------------------------------
Norman Gray : Physics & Astronomy, Glasgow University, UK
http://www.astro.gla.ac.uk/users/norman/ : www.starlink.ac.uk
|