On Thu, Jun 12, 2008 at 11:50 AM, Martin Pitt wrote:
> geode debdiff:
> - why is it necessary to move around the -amd symlink in the SRU, and thus mess with new Conflics/Replaces:?
To make it possible to completely purge the symlinks without purging
the driver. This is necessary to fix operation in LTSP.
> - what is the rationale for dropping all the DEB_* settings in debian/rules?
Not dropped. They were accidentally duplicated before.
> Judging by the noisy changelog diff, this is really a backport of the
> latest version in intrepid/sid, as opposed to a minimally intrusive
> patch to fix just this issue. I guess the important bits are the
> additional ID in geode.ids and moving "script" from -amd to -geode
> perhaps?
geode.ids and making the symbolic links removable.
> Please supply minimal debdiffs which can be reviewed and avoid any
> unnecessary change, which might just cause trouble later on.
All the above changes *are* necessary. Trying to avoid them earlier is
what resulted in 2.1 not fixing anything.
> nsc debdiff: same like above, but amplified by 10. Adding new packages
> (-dbg), changing the build system, changing config.{guess,sub} are
> things which are inappropriate for an SRU.
I'm sorry, but I cannot work with the XSF scripts.
On Thu, Jun 12, 2008 at 11:50 AM, Martin Pitt wrote:
> geode debdiff:
> - why is it necessary to move around the -amd symlink in the SRU, and thus mess with new Conflics/Replaces:?
To make it possible to completely purge the symlinks without purging
the driver. This is necessary to fix operation in LTSP.
> - what is the rationale for dropping all the DEB_* settings in debian/rules?
Not dropped. They were accidentally duplicated before.
> Judging by the noisy changelog diff, this is really a backport of the
> latest version in intrepid/sid, as opposed to a minimally intrusive
> patch to fix just this issue. I guess the important bits are the
> additional ID in geode.ids and moving "script" from -amd to -geode
> perhaps?
geode.ids and making the symbolic links removable.
> Please supply minimal debdiffs which can be reviewed and avoid any
> unnecessary change, which might just cause trouble later on.
All the above changes *are* necessary. Trying to avoid them earlier is
what resulted in 2.1 not fixing anything.
> nsc debdiff: same like above, but amplified by 10. Adding new packages
> (-dbg), changing the build system, changing config.{guess,sub} are
> things which are inappropriate for an SRU.
I'm sorry, but I cannot work with the XSF scripts.