yum flavors vs/ fc1, fc2, fc3...infinity

[Jeff Vian]

On Sat, 2004-07-17 at 20:47, Jeff Ratliff wrote:
> On Sun, Jul 18, 2004 at 01:55:59AM +0100, Timothy Murphy wrote:
> > Jeff Ratliff wrote:
> > 
> > > I'm sure the developers would have liked to test on a wider range
> > > of machines, but testing is voluntary. You have yet to propose a
> > > solution to this problem.
> > 
> > I did actually make a suggestion,
> > namely that it should be ensured as far as possible
> > that the test distributions could be upgraded to the final distribution.
> > 
> > I would run the test distributions if that were the case,
> > and I imagine there are many more like me.
> > 
> I've updated over 94 packages since I installed FC3T1 two
> days ago, including 2 new kernels. Who knows how much cruft and
> broken stuff will be there by October? It just doesn't seem possible
> to guarantee a test release will be upgradable without seriously 
> hampering testing. I plan on wiping the parition, installing FC2 
> and upgrading to FC3test with each test release, so that anaconda
> can be tested. If the installer was sufficiently tested, things
> like the dual-boot bug probably wouldn't have gotten out the door. 
> 
> My guess is that most people will be upgrading from FC2 to FC3, 
> and that's what testing is trying to target. 
> 

Although my experiences are not exhaustive in context, they are
representative.

IMHO the problem with anaconda and the dual boot bug is not really a bug
at all, but rather it identifies those instances where the BIOS use of
LBA is broken.  

It seems to me that it applies to any machine I have tested it on where
the BIOS fails to display the correct LBA mapping of the CHS.  The ones
I have seen fail have displayed the physical CHS of the drive, and have
also said LBA was enabled.  As we all should know, LBA does not
display/use the physical CHS but instead uses the logical mapping.  Thus
if bios is set to auto configure LBA and the device still shows the
physical mapping, then it is not correctly using/displaying LBA and is
likely to experience the dual-boot failure with an FC2 install (and in
some cases may not even install FC2 correctly). 

It is impossible for testing to even identify that failure unless
someone happens to do a dual-boot install on one of the affected
machines. This is likely why it was not well known before FC2 was
released.

Upgrades would be nice if it can be made smooth AND if all needed
configuration changes can be handled.