Don't write /etc/rpm/macros. (#650490)
 

On Mon, 8 Nov 2010, Bill Nottingham wrote:


Chris Lumens (clumens@redhat.com) said:

This branch == rhel6-branch, if I wasn't obvious. Although, I
wouldn't argue against it on master either.


Though, I'd prefer to remove all the callers and the method entirely.


In looking, this function is called from /usr/bin/anaconda, and
yuminstall.py:doBackendSetup(), both to write the config for the
yum/rpm instance that anaconda uses. That implies that anaconda currently
sets it to prefer 32-bit binaries if a binary exists for both arches in
the transaction. Given that we set multilib_policy=best, and we compose
with ppc64 as primary, this shouldn't be an issue in any default
installation. However, if someone manually adds <foo>.ppc to their
package set (via kickstart, or whatever) this would change the behavior
they see.

Do we still want to change this in released RHEL? While it's a behavior
change, I think the current behavior (where yum preferred 64-bit packages,
but you got 32-bit binaries if you installed both) is a bit nonsensical.


I agree that the current behavior does not make sense, but since RHEL-6 is
already going, I'm going to say we hold the line and not make this change
unless it comes in as a bug. We've seen too many instances where things we
think should be cleaned up end up surprising way too many users. We should
change master as clumens indicated, but keep the rest of rhel6-branch
functionality as-is and only make the change necessary for this bug report.

--
David Cantrell <dcantrell@redhat.com>
Red Hat / Honolulu, HI

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

Don't write /etc/rpm/macros. (#650490)
 

David Cantrell (dcantrell@redhat.com) said:
> >In looking, this function is called from /usr/bin/anaconda, and
> >yuminstall.py:doBackendSetup(), both to write the config for the
> >yum/rpm instance that anaconda uses. That implies that anaconda currently
> >sets it to prefer 32-bit binaries if a binary exists for both arches in
> >the transaction. Given that we set multilib_policy=best, and we compose
> >with ppc64 as primary, this shouldn't be an issue in any default
> >installation. However, if someone manually adds <foo>.ppc to their
> >package set (via kickstart, or whatever) this would change the behavior
> >they see.
> >
> >Do we still want to change this in released RHEL? While it's a behavior
> >change, I think the current behavior (where yum preferred 64-bit packages,
> >but you got 32-bit binaries if you installed both) is a bit nonsensical.
>
> I agree that the current behavior does not make sense, but since RHEL-6 is
> already going, I'm going to say we hold the line and not make this change
> unless it comes in as a bug. We've seen too many instances where things we
> think should be cleaned up end up surprising way too many users. We should
> change master as clumens indicated, but keep the rest of rhel6-branch
> functionality as-is and only make the change necessary for this bug report.

'the change necessary for this bug report'... the change for the bug report
would be for anaconda to not write the file on the installed system. If we
only changed it there, then the behavior during install and post-install
would be different. I'm not sure that's better.

Bill

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

Don't write /etc/rpm/macros. (#650490)
 

> 'the change necessary for this bug report'... the change for the bug report
> would be for anaconda to not write the file on the installed system. If we
> only changed it there, then the behavior during install and post-install
> would be different. I'm not sure that's better.

Yeah, that kind of maddening inconsistency is the sort of thing we
usually try to avoid. It looks like the choices are:

(1) Don't do anything on rhel6-branch and live with the consequences for
the next 500 years.

(2) Make the change on rhel6-branch and break some people's
expectations. It only hits on ppc, so the damage is at least a little
contained. Perhaps we can make the change and let everyone know why
we're doing it well ahead of time so no one's surprised?

- Chris

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

Don't write /etc/rpm/macros. (#650490)
 

On Mon, 8 Nov 2010, Chris Lumens wrote:


'the change necessary for this bug report'... the change for the bug report
would be for anaconda to not write the file on the installed system. If we
only changed it there, then the behavior during install and post-install
would be different. I'm not sure that's better.


Yeah, that kind of maddening inconsistency is the sort of thing we
usually try to avoid. It looks like the choices are:

(1) Don't do anything on rhel6-branch and live with the consequences for
the next 500 years.

(2) Make the change on rhel6-branch and break some people's
expectations. It only hits on ppc, so the damage is at least a little
contained. Perhaps we can make the change and let everyone know why
we're doing it well ahead of time so no one's surprised?


If this is only on ppc, I'm less worried about making the change. And now is
better than later, so let's go with it and hope no one freaks out.

--
David Cantrell <dcantrell@redhat.com>
Red Hat / Honolulu, HI

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list

Don't write /etc/rpm/macros. (#650490)
 

David Cantrell (dcantrell@redhat.com) said:
> >(2) Make the change on rhel6-branch and break some people's
> >expectations. It only hits on ppc, so the damage is at least a little
> >contained. Perhaps we can make the change and let everyone know why
> >we're doing it well ahead of time so no one's surprised?
>
> If this is only on ppc, I'm less worried about making the change. And now is
> better than later, so let's go with it and hope no one freaks out.

OK, pushed for rhel6-branch. Bigger patch sent for master.

Bill

_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@redhat.com
https://www.redhat.com/mailman/listinfo/anaconda-devel-list