15/11/2008
Gentoo’s epic phail
As some people already know I’ve joined the army since 2 months ago, this makes it somewhat difficult for me to keep up with the latest updates for every machine I use.
Today I tried to upgrade a machine running stable (x86) Gentoo Linux after more than 15 days since the last upgrade and I got confronted with an epic Gentoo failure. The problem is clearly described on Odi’s blog. While trying to update e2fsprogs you have to uninstall the old version of them (so far so good), remove sys-libs/ss (that’s also acceptable) and remove sys-libs/com_err which is a dependency for MANY MANY progs, wget among them. So when you remove all these packages and try to install e2fsprogs-libs, wget can’t work anymore due to the missing libcom_err.so.2 file, so you can’t download the updates! You can’t open a new ssh connection to the machine either, so you can’t sftp the libcom_err.so.2 file from another machine! I had success by placing libcom_err.so.2 on a nfs share in another machine, mounting that share from the first machine and trying to re-emerge e2fsprogs-libs since wget could then work.
You can read more about the bug on Gentoo’s bugzilla, here and here.
a) Since portage is NOT ready to handle these kind of situations (according to bugzilla) why did the maintainers mark e2fsprogs-libs as stable ? Why didn’t they do ANY testing at all on the stable branch ?
b) Why wasn’t there a warning/alert/whatever on Gentoo’s website ?
c) It’s been more than 10 days since the problem first appeared and there hasn’t been any official solution about it, either by portage upgrade or by package masking.
d) Should I have googled or searched the forums before upgrading ? Possibly yes…but Gentoo didn’t have such upgrade problems before. I could accept it if problems on Gentoo were created by an upstream ABI breakage like dev-libs/expat had not so long ago, but this looks to me as a totally Gentoo related problem and not an upstream one.
I award Gentoo and e2fsprogs-libs maintainers a sad trombone. Sorry people but this is an epic failure. You deserve it.
Filed by kargig at 17:23 under Gentoo,Linux
16 Comments | 8,180 views
Flexibility is a nice thing, but some times, if you operate on such low levels it may also cause problems
The dependency issues are a major pain in the butt, especially if you update in longer intervals (than the maintainers would like you to). I still remember an older Redhat installation i had, which tried to rpm update about 400 packages, some of them VERY core libraries…
Anyhow, now that you are in the army (and therefore you are a man), you can ditch gentoo and start working with manly operating systems (call me macosx)
Very sorry to hear your case. If you could, try to use the new portage (2.2) it handles all this things much better and so it’s much less prone to breakage than any other version I’ve used until now.
totally agree with you kargig
works for portage with ~x86 flag
@Lukas: I guess you are 100% correct about ~x86, but that’s not how things should have normally been. Developers should first make sure that it works 100% correct on stable and then care about unstable. People trust the stable branch to be “stable” and not to have problems (like making their PC unbootable). If gentoo can’t keep up anymore with 2 branches that’s a hell of a different story and we should be told by the council to switch from stable to unstable (and then move of course to another more stable distro…).
I’ve been a looooong time supporter of Gentoo and I’ve even given presentations about it in Greece but it was never that messy before. The sooner people realize how badly it’s falling apart the better. I hope it gets better soon 🙂
This does remind me of a time long ago when the LVM maintainer decided that a critical tool should no longer be statically linked, resulting in a failure to mount any of my partitions (save a minimalistic root) after a reboot, since the respective dynamic libs lied in the (LVM) /usr logical volume.
This was the cause for 2-3 lost hours back at the time and I dumped gentoo after this incident. Prior to it I had frequently argued (with GKeramidas mainly) that Gentoo is not an immature distro but rather has some immature maintainers for some packages. Problem is that when immature maintainers control that important packages (like e2fsprogs you brought up or LVM tools) the end result is the same.
If somebody is watching gentoo-user list he must already know that this error was heavily discussed . It is ok now and then to have such problems. Developers are just people and they are allowed to do some mistakes. What we should do is to provide accurate information on gentoo bugzilla to assist them.
Furthermore i didnt face this problem on both of my Gentoo machines (stable and unstable one). What I did was to uninstall the older version ( and the e2fs-libs ) and install the new one ( as far as I remember )
Anyway 🙂
If somebody is watching gentoo-user list he must already know that this error was heavily discussed.
Replies like these are a good indication that Gentoo is (still) an immature distro. I don’t have the time or want to monitor any ${distro} or ${o/s} mailing list or forum, I want my computer to just work, whether it runs Vista, XP, Gentoo, Ubuntu, Debian, Solaris, you name it. Running the equivalent of “apt-get update && apt-get upgrade” on such systems and being certain that the core o/s and major daemons and services (i.e. SSH, Apache) will continue to function smoothly certainly falls within “just works” definition.
Developers are just people and they are allowed to do some mistakes.
There are mistakes and mistakes. A minor package ending up broken is acceptable. Your system catching fire if you run an experimental kernel is probably also acceptable. Things occasionally breaking up badly in whatever unstable branch your O/S or distro use is OK. Running a supposedly stable O/S branch and having filesystem utilities breaking is not.
Unfortunately I have to agree with mperedim that I don’t want to monitor mailing lists to run a OS on my desktop. Gentoo is not immature, it’s just in the middle of a crisis, developers have lost focus and there’s nobody to pull any strings.
All distros have problematic packages from time to time, let’s not forget that Ubuntu, which supposedly is the most user friendly distro had posted twice an update that broke Xorg, and breaking Xorg for an ubuntu user is probably worse than a problematic kernel update for a Gentoo user 😛
Arch linux had also once issued a kernel update that make any machines that updated unbootable.
Shit happened, happens and will happen.
And the list goes on and on…
What I think that matters most is the way that the distro handles the problem, and Gentoo still ignores it. No public announcement yet, just some mailing-list and bugzilla posts which is certainly not enough.
Yup, I ran into it too. I was trying to figrure out why json and curl are not running on my php and decided to update to make sure I hadn’t run into a problem. Someone else built this machine. Since the aforementioned blockages were there I removed them per the manual.
Woops.
Now I had a machine in a Tier-1 central office 15 miles away that I couldn’t get to that couldn’t reach out via *any* network facility.
Fortunately A) I was still logged in and B) I had access to another gentoo machine. Since I couldn’t scp, I used uudecode on the second machine (look it up kids) and copied the text via cat – >uufile to the 1st machine. Naturally the 1st machine was not installed by an old timer (me) and didn’t have uudecode, so I had to use perl to write it. Run the perl, get the libcom_err.so.2 and replace.
How to deal with my rage at getting screwed over by gentoo and the guy who installed this in the first place yet again, I have yet to figure out.
idiots.
Gentoo sucks. I had told you. Do you understand why another distro needs 2 development branches now?
Oh, and I see you use a Mac. What next? Windows?
I’ve said exactly what sucks on Gentoo. If you just commented here to start a flame, tough luck. I am so over distrowars…grow up.
Oh and I am using windows from time to time too. What’s the problem with that ? :/
I ran into the same problem, and I’m getting sick of having to jump through hoops every time I update my system. Since I liked a certain app that conflicted with more poorly done app that was part of a large meta-package (KDE IIRC), I would have to make sure certain things were removed before updating other things, and that everything got installed in a certain order. Not to mention, for the last 6+ months there would be packages where I would get compile errors. They weren’t simple things like libxyz missing, but things like undeclared variables or missing/extra symbols like ;, { or }. I could understand if I were trying to run packages that were hard masked, but these were marked stable!
At least I had Firefox open when this problem came along, so I could download the packages I needed. The problem that ruined my system was a rather easy fix compared to most. I still can’t figure out how to re-install media-sound/jack-audio-connection-kit-0.103.0
P.S. Mailing lists for an OS should be for the OS developers and maintainers.
Debian ftw! 😉
😛
Till they fix this…
euler dionyziz # cat /etc/portage/package.mask
=sys-fs/e2fsprogs-1.41.2
I had the same problem few weeks ago (yep, I have got a big delay in upgrades) and I couldn’t do nothing with the machine (it was in different country). But I found info about busybox. This app doesn’t need libcom_err (and it was installed on the machine), so I was able to download missing files.