Bug 104152

Interesting. Not supposed to ever get to that block with an
empty list of skipped packages (which seems to be whats
happening).

Couple quick run thoughs doesnt show an easy duplication.
Any thing slightly odd of normal about that run? lots of
going backwards or something?

No going backwards.  It's right after the bug reported in #104170, though.
I saw it again on the other severn system I'm running just now.

It appears to be related to ##104170, which should be
fixed now. (more specifically, with the versions that
fail to restart and then attempt to jump to a somewhat
bogus saved state). The state info is bogus causing
the error show. Should not happen with the fix for
#104170. 

Let me know if you see it again with 3.9.26 or newer. 

Aha, I finally paid enough attention to notice that this happens when a
"Packages Flagged to be Skipped" window pops up.  It has zero packages
listed in the window.

Incidentally, the last words on the screen, "select its checkbox", only show
the tops of the characters unless I resize that window larger.

I should explicitly note that comment #4 is using 3.9.26-2

*** Bug 104234 has been marked as a duplicate of this bug. ***

Hmm, when I ran 3.9.26-2 w/o the rawhide and beta keys imported (just up2date,
w/o the --nosig) I got the empty skip list and the traceback

Once I imported the rawhide and beta keys, then ran again (again w/o the
--nosig), I didn't get the empty skip list, and I didn't get the traceback

Not sure if it's coincidental or not, though ;-)

re comment #4, did the screen go directly to the
"Skipped Packages" or the more normal up2date?

At least for me, it went first to empty Skipped Packages, then into the normal
up2date package selection

When it happend to me in bug 104234, it went directly to the "Skipped Packages"
list with nothing listed.

okay, going directly to "skipped packages" (ie, no welcome screen, nothing)
is the behaviour when it restarts after updating itself (it caches the
already computed avaiable packages and such). 

So were this "skipped packages" screen errors after a bogus restart
with <= 3.9.25 (ie, a restart failed, then the next time up2date was
ran, you got the "skipped packages" directly and then the error) (option 1)
or did it all happen from one "up2date" invocation (up2date starts
up, updates itself, restarts correctly, goes to the skipped
package screen, and throws the traceback)? (option 2)

Or of course, None of the Above?

Note that versions prior to 3.9.26 when ran from a user account, will
fail to update/restart themselves properly. This leaves a bad state
file (causing the jump to empty packages). When up2date starts
up, it reads the state file, jumps ahead, then deletes the state
file (which is why you will only see this behaviour once).

Does that make sense? Any one seen this fail in a matter that
that doesnt seem to jive with?

(option 2) for me, and this was ran from the root account.

Mine was option 1 as a non-root user -- ran older version, updated up2date,
tried to restart, failed, restarted manually from command line, went straight to
skipped and non-fatal traceback, subsequent runs from command linewere
traceback-free

hmm, okay. The plot thickens. 

Now just to figure out why I can't reproduce this with 3.9.26

Thanks for the info

I had exactly the same experience as kaboom.

*** Bug 104244 has been marked as a duplicate of this bug. ***

I belive this is fixed in 3.9.27. Or at list, after reproducing
the errors mentioned above with 3.9.26, the fixed in 3.9.27 seem
to make them go away for me.

Upgrading 3.9.26 to 3.9.27 should work fine, and then 3.9.27 should
be fine when it restarts. 

*** Bug 103185 has been marked as a duplicate of this bug. ***

This looks like it's fixed. I could not duplicate it going from up2date-3.9.26-2
to up2date-3.9.28-2