Why was a kernel-2.6.34 pushed to updates that had un-addressed bugs. / Every OS sucks!

[John Morris]

On Mon, 2010-09-06 at 23:05 -0400, Chuck Ebbert wrote:
> On Thu, 02 Sep 2010 18:23:01 -0500
> John Morris <jmorris at beau.org> wrote:
> 
> > In my case I reported #573135 back in March and stopped taking kernel
> > updates. In another month or so I'll boot a live USB stick of F14 and
> > see if the bug was fixed and just didn't get closed.  Then it is either
> > suck it up and run without security fixes or jump distros.
> > 
> 
> And in the meantime these patches went into the F12 kernel via
> 2.6.32-stable, but you weren't even checking the updates:

Sorry everyone, time to vent.

Bah.  This is why I just blocked kernel updates in the first place.
Tried updating.... now things are worse due to a bad combination of
Fedora policy multiplied by my own stupidity.  I KNEW you were supposed
to make darned sure you were running your favorite kernel before letting
yum loose. (Don't ask, long story)  That was mistake number one.  Number
two came when I looked to the backup server and found to my horror I
fumbled that too, / and /home safely backed up but no /boot!  Double
crap!

So now I lost the only kernel package where everything worked.  And of
course Fedora doesn't have it anymore.  You can pick the original
package or the current update.  Triple crap!   If anyone has a pointer
to kernel-2.6.31.12-174.222.x86_64.rpm I'd really appreciate it!
Google, rpmfind, etc. all come up blank as did manually poking around on
the Fedora mirrors.

Ya would think there could be ONE site where someone could find old
packages.  Figuring out exactly where an unnoticed regression appeared
would be a lot easier with a complete archive of updates.  Ban direct
access by yum using the user-agent string to prevent people from abusing
it if needed.  Seriously, if there really isn't any complete archive of
Fedora updates I'd be willing to host it.  Assuming the wizards who wear
red headgear even have a complete dataset.

With the current kernel package the ACPI error is gone but it still
doesn't survive a suspend/resume so apparently there is more than one
regression involved.  The only kernel package I can still find that
sorta suspends breaks Network-Manager so that after resume a manual kick
of it is usually required.  At least I think that is what is happening,
it might have been an update to NM a couple days before because I think
I remember having to kickstart NM once before I tried the kernel update.

Now for the ranting part.  Not so much about the above bug, it is after
all a fairly exotic one involving power management.  But about the
rising misfeature problem in general.

And of course Network-Manager isn't optional anymore.  Oh no, you can't
just revert to the original networking subsystem that actually worked,
years ago, better for everything except WiFi.  Onwards into the glorious
future!  And no Citizen, totally replacing the networking subsystem was
easier than adding WPA/WPA2 support to the old one.  Never question the
developers!  Yes it has been several years now and NM still doesn't do
bridging (well if at all) multiple static IPs per interface or PPPoA but
we know what is best for you.  So you obviously didn't need those
features.  What?  Virtualization really needs bridging you say?  Fie on
you!

Some days lately I actually consider just saying screw it, boot the Win7
partition and try the Cygwin path to UNIX happiness.  That multiplatform
OpenSource migration path goes both ways... Folks say OO.o, FF, Putty
and The Gimp are pretty stable on Windows lately.. which is more than I
can say for FF on Fedora.  Is there a replacement for XMMS?  The last
Windows I used for more than a hundred hours was Win95 so really don't
know what it is like now.  Want to follow the GNU Testament but the
Linux desktop has been becoming less stable over the last few years and
everyone says Win7 is actually getting fairly reliable at long last.
But I have booted it a couple of times on this Thinkpad... enough to
know the multi-monitor support in Win7 is as dodgy as Fedora's and that
unlike Fedora I can't fix the breakage with a bash script bound to a
hotkey.  When Windows breaks you are just outta luck.  No
bugzilla.microsoft.com.

Is this why so much malevolent glowing fruit at conferences?  Is it a
warning sign when so many open source devels stop eating the dog food?

And to make this more than just a Fedora vs Windows rant.... spent this
afternoon helping a coworker try Ubuntu only to discover the automounter
has apparently been broke for months; the discussion on the bug has,
after months, come really close to a solution but no actual update has
issued yet.  Really?  The automounter known to be 100% broken for
MONTHS?  Not broken in certain corner cases, not broken for some users,
100% broken for 100% of users.

Or check this one.  Debian's dial up networking support is currently
busted.  Ok, not a lot of people still use it (Except for the 299
Windows using library patrons who dialed into our Portmasters in
August.) and I might just be going senile but last month I tried a fresh
Debian[1] + a genuine US Robotics V.Everything external and got nowhere.
Dial, carrier and then fail to negotiate a PPP link.  Logs didn't say
anything useful on either side.  That was the first time I failed to get
a dialup link to work since Yggdrasil Linux and Trumpet Winsock on
WfW3.11.[2]  Finally had to give up that day and move on, but hope to
get back to it; turn on debugging and figure it out so I can at least
file a proper bug report.  Perhaps someone will care.

Folks, we have a problem.  If there isn't a "come to Jesus" moment real
soon the Linux/GNU/X desktop is about to collapse under this breakneck
development pace.  For the first decade almost every release was a major
improvement regardless of which distribution.  Now it is mostly sideways
motion with change for change's sake, pointless rewrites instead of
repairing, breakage and misfeatures almost as common as actual
improvements.

BREAKING CORE SUBSYSTEMS FOR MONTHS/YEARS IS NOT ACCEPTABLE.  Rawhide,
yes.  Alphas only by accident.  Never in a shipping system.
NetworkManager and PulseAudio might be the poster children for
multi-year breakage but the mindset is fast becoming mainstream.  It is
really hard to get work done if at any point in time at least one major
subsystem (and lately more than one) is unreliable to the point of being
unusable by mortals.

A minor security issue that could in theory allow an exploit, if the
right corner case can be abused, will cause a full stop and a rush to
fix it.  Every serious distribution keeps a security team ready to
spring into action on a moments's notice.  PulseAudio left countless
real users lacking the skills to deal with it with no working audio for
years and nobody in a position of authority saw a problem important
enough to even postpone the work to bind PA so deeply into the system
that turning it off was no longer an option.  Am I the only one who has
ever wondered if there just might be a problem in setting priorities?

[1] Why Debian?  Because Fedora decided a 1% speed bump was worth
ditching support for the Via EPIA series.  Or someone just got tired of
explaining for the Nth time that a .386.rpm compiled with the right
switches was just as good as those .[56]86.rpm packages distro Y's
fanboys were always holding up as an example of superiority.

[2] And yes you can get the heck off my lawn now.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
Url : http://lists.fedoraproject.org/pipermail/test/attachments/20100908/6222651d/attachment.bin