fc3test upgrade to fc3 final

[Petr Fischer]

Hi,

can I upgrade from Fedora Core 3 Test 3 to Fedora Core 3 Final _without_
_problems_ in future with Yum?

I can't wait for FC3 final release and FC2 kernel has some problems with
my ASUS board.

Thanks, pf

[Elliot Lee]

On Fri, 29 Oct 2004, Petr Fischer wrote:

>  can I upgrade from Fedora Core 3 Test 3 to Fedora Core 3 Final
> _without_
>  _problems_ in future with Yum?
>
>  I can't wait for FC3 final release and FC2 kernel has some problems with
>  my ASUS board. 
(a) Test releases are meant for testing only (which means you can't count 
on them to remain useful when final releases come around)
(b) upgrading via yum is not supported. Sometimes it works, sometimes it 
doesn't.

With those caveats in mind, an upgrade via anaconda *probably* will work
(it did for me).
-- Elliot

[Andrew W. Donoho]

On Oct 29, 2004, at 15:47, Elliot Lee wrote:
>  (a) Test releases are meant for testing only (which means you
> can't 
>  count
>  on them to remain useful when final releases come around)
>  (b) upgrading via yum is not supported. Sometimes it works, sometimes 
>  it
>  doesn't. 

And what should be my expectation for rc releases? (I am planning on 
moving a test machine from FC2 to FC3rc3.) Since this is a release 
candidate, presumably yum will upgrade any changing components between 
rc3 and final?



>  With those caveats in mind, an upgrade via anaconda *probably* will 
>  work
>  (it did for me). 
There is always that...

Thanks,
Andrew


____________________________________
Andrew W. Donoho
awd(a)DDG.com, PGP Key ID: 0x81D0F250
[phone] (o), [phone] (m)

[Jef Spaleta]

On Fri, 29 Oct 2004 16:01:45 -0500, Andrew W. Donoho <awd(a)ddg.com&gt; wrote:
>  And what should be my expectation for rc releases? (I am planning on
>  moving a test machine from FC2 to FC3rc3.) Since this is a release
>  candidate, presumably yum will upgrade any changing components between
>  rc3 and final? 
presumably? there could be a last minute package downgrade between rc
and final could interfere with upgrading from an rc in this manner.

-jef

[Timothy Murphy]

On Friday 29 October 2004 23:02, Jeff Spaleta wrote:

>  presumably? there could be a last minute package downgrade between
> rc
>  and final could interfere with upgrading from an rc in this manner. 
If you found there was a bug in version 7,
and you wanted to revert to version 6,
would it not be simpler to call it version  8?

-- 
Timothy Murphy  
e-mail (<80k only): tim /at/ birdsnest.maths.tcd.ie
tel: [phone]
s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland

[Per Bjornsson]

On Sat, 2004-10-30 at 17:36, Timothy Murphy wrote:
>  On Friday 29 October 2004 23:02, Jeff Spaleta wrote:
>  
>  > presumably? there could be a last minute package downgrade between rc
>  > and final could interfere with upgrading from an rc in this manner.
>  
>  If you found there was a bug in version 7,
>  and you wanted to revert to version 6,
>  would it not be simpler to call it version  8? 
That definitely depends. If a packaging fix is applied, then yes, of
course you bump the RPM release tag. However, look at what happened to
Evolution in FC2. For a while during the test releases, the development
series Evolution 1.5.x was in Rawhide (and in FC2t1 I believe). However,
once it became clear that the stable version derived from the 1.5
series, which is the 2.0 series, wouldn't be done in time for FC2, it
was backed down to the 1.4 series. As far as I can see, there's just no
really sane way to recover from this; an epoch bump would make the
downgrade happen automatically, but that is a somewhat draconian measure
to fix a problem in the beta stage, and leaves you with dealing with the
consequences forever. At this stage, to me the decision to let the beta
testers just deal with it seems perfectly sane. Users of test releases
are supposed to know what they're doing, or at least live be able to
with their decisions.

Also, the kernel version went backwards just before FC3t3; as far as I
could tell what happened was pretty much that there were some pretty
serious concerns about the later kernels (mostly SELinux related). There
wasn't really enough time to fix everything correctly so by far the
simplest solution was simply to dump the "last known good" kernel into
FC3t3.

Cheers,
Per

[Elliot Lee]

On Fri, 29 Oct 2004, Andrew W. Donoho wrote:

>  
>  On Oct 29, 2004, at 15:47, Elliot Lee wrote:
>  > (a) Test releases are meant for testing only (which means you can't 
>  > count
>  > on them to remain useful when final releases come around)
>  > (b) upgrading via yum is not supported. Sometimes it works, sometimes 
>  > it
>  > doesn't.
>  
>  And what should be my expectation for rc releases? 
The same... "It ain't over till it's over."

However, the set of changes from rcN -> final will be a lot smaller, so if
you like taking risks and know how to manage them by resurrecting a broken
system yourself, the odds are highly in favor of a successful upgrade.

-- Elliot

[Timothy Murphy]

On Monday 01 November 2004 18:00, Elliot Lee wrote:

>  > And what should be my expectation for rc releases?
>
>  The same... "It ain't over till it's over." 
This argument about the danger of upgrading is often given,
but I have never seen the opposite danger of a new installation spelt out.

In my experience, the kernel in new installations frequently does not work -
I haven't seen any kernel since Redhat-8.0 which ran
on my not very unusual SCSI-only machine.
I've also had problems with the kernels in new installations
on my Sony Picturebooks.

Perhaps I am atypical, but I have found
that the chances of a new installtion running
are much lower than the chances of an upgrade running -
because I can choose an old kernel I know to work with an upgrade.

Also of course if an upgrade does not work 
one can always try a clean installation,
while the converse is not possible.


-- 
Timothy Murphy  
e-mail (<80k only): tim /at/ birdsnest.maths.tcd.ie
tel: [phone]
s-mail: School of Mathematics, Trinity College, Dublin 2, Ireland