Re: [PATCH] Use device ids instead of paths for action pruning. (#500808)

[David Lehman <dlehman redhat com>]

On Thu, 2009-05-28 at 11:21 -0400, James Laska wrote:
> On Thu, 2009-05-28 at 11:15 -0400, Chris Lumens wrote:
> > > Both look good to me, I wonder if this is F-11 material though.
> > 
> > Yeah this is really late.  Unfortunately the renumbering thing is
> > hitting us in a couple really major ways:
> > 
> > - The protected partitions get lost if you delete earlier partitions.
> > - If you create some new partitions, then delete them, then create
> >   new ones.
> > - If you delete logical partitions and then the extended partition,
> >   getting the sort order incorrect.
> > 
> > If we don't fix this, we are going to see some serious problems in F11.
> > The problem is that we are fundamentally changing how partitions are
> > referenced in a couple places and that really needs more testing than we
> > are going to be able to give it.
> 
> I suspect, the above scenarios are more likely in custom partitioning
> use cases (which seems common enough).  

It also breaks autopart if the initial disk layout contains multiple
logical partitions.

> 
> A few follow-up questions as it relates to F11 ...
> 
>       * If we don't fix this, are there workarounds we can document in
>         https://fedoraproject.org/wiki/Common_F11_bugs 

The big problem is that autopart will be broken for setups with two or
more preexisting logical partitions on a given disk, which is pretty
common.

We could post an updates.img, but most folks won't find it until it's
too late.

It might at least solve the autopart/clearpart case if we changed
clearPartitions to iterate over the partitions in reverse order, FWIW.

>       * If we do fix this, what other partitioning areas would be
>         impacted?  Would this warrant a reset of all F11 partitioning
>         testing?

This impacts the sorting of partitioning actions, so it plays into every
partitioning scenario. It has little effect on anything else.

I cannot say it necessarily warrants a complete reset, but it will
certainly require thorough testing across a variety of scenarios with
some emphasis on multiple preexisting logical partitions.

>       * Should this bug be on the F11AnacondaBlocker list? 

Probably.

Dave

> 
> Thanks,
> James
> _______________________________________________
> Anaconda-devel-list mailing list
> Anaconda-devel-list redhat com
> https://www.redhat.com/mailman/listinfo/anaconda-devel-list