Re: Massive testing breakage; unstable: no text in GnuCash or E16
On Thu, 2006-05-18 at 17:10 +0200, Gabor Gombas wrote:
> On Thu, May 18, 2006 at 09:22:10AM -0400, Adam C Powell IV wrote:
>
> > So I tried starting E, then gnome-settings-daemon, then when I run
> > gnome-panel, bam! X crash, framebuffer problems, and kernel panic on
> > warm reboot.
>
> I'm not sure I understand correctly. If the kernel paniced _after_ the
> reboot, that may be caused by your BIOS not initializing some device
> correctly.
Right, sorry if I was unclear, the panic happened after the reboot.
This is a BIOS issue for sure, but has only happened on this machine
since upgrading to 2.6.15 (and now 2.6.16), so something new in the
kernel is tickling the BIOS bug. Not hoping for resolution on this
list, just trying to indicate the depths of brokenness of the stack in
etch.
> > FWIW, a friend here at MIT has switched to KDE, and the family of a
> > colleague at Northeastern is considering dropping GNU/Linux entirely
> > (much to my colleague's chagrin) because of the breakage in this
> > transition. Next time we upgrade, can we please stage it through
> > experimental, like the very smooth 2.6 and 2.8 transitions? This
> > piecemeal approach has caused a ton of problems, and yes testing and
> > unstable have those names for a reason, but avoiding preventable
> > breakage would be most welcome.
>
> I'm tracking unstable almost daily on both an i386 and an amd64 machine
> and I did not have any serious issues with the X transition (and I've
> switched to Xorg 7 while it was still in experimental on i386). Btw., I
> don't consider editing a couple of config. files after an upgrade an
> "issue"; that's part of the normal sysadmin tasks.
Editing config files? Not sure where you saw that in this thread. So
you didn't have problems with the sudden disappearance of .la files in
libxcursor and libxrender? You can run gnucash in unstable? Pray tell.
> I'm also using E and never had any issues with it for many years now.
>
> If you're using testing or unstable you should never ever just blindly
> do an upgrade (much less a dist-upgrade). You should _ALWAYS_ examine
> what packages are going to be changed, read their changelog, examine
> what packages are going to be held back and why. Look at any packages
> that are important for you twice. If you're not sure about a package,
> then do not upgrade that package. Checking the relevant Debian mailing
> lists for potential problems _before_ the upgrade also helps a lot.
I think you're barking up the wrong tree, or perhaps preaching to the
choir. I was the only one who reported this to the mailing lists. And
if the changelogs had "CAREFUL! This upgrade will break existing
gnome-panel, and interacts badly with X -- which in turn is tickling
kernel bugs!" then I clearly would have avoided it. :-)
> You should also be aware that due to the rules of testing, sometimes
> testing can be more broken than unstable. Don't beleive the names.
>
> With these simple rules you can minimize unpleasant surprises greatly.
Indeed. Unfortunately I only learned of my colleague's family upgrade
from sarge to etch after the fact, and said, "That was not smart". :-(
My research group remains on sarge for this reason, I only upgraded my
own machines in a desperate search for a fix to GNOME bug 320249, which
I reported...
Cheers,
-Adam
--
GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6
Welcome to the best software in the world today cafe!
http://www.take6.com/albums/greatesthits.html
Reply to: