| View previous topic :: View next topic |
| Author |
Message |
nein
Guru
Joined: 14 Nov 2003
Posts: 346
Location: Spain
|
 Posted: Sun Dec 05, 2004 6:46 pm Post subject: |
 |
|
| Quote: |
don't downgrade glibc! applications linked against the new one will stop working after downgrade. there is/was some thread about it in the past, even marked sticky, I think...
|
Already done it and no problems so far. I will emerge the latest version again once this problem with the relocation error is completely solved.
Now I can at least work a bit without having to prelink |
|
| Back to top |
|
 |
jkt
Retired Dev
Joined: 06 Feb 2004
Posts: 1250
Location: Prague, Czech republic, EU
|
|
| Back to top |
|
|
John-Boy
Guru
Joined: 23 Jun 2004
Posts: 442
Location: Desperately seeking moksha in all the wrong places
|
| Posted: Sun Dec 05, 2004 7:09 pm Post subject: |
|
|
| jkt wrote: | | nein wrote: | I was about to downgrade glibc but noticed that each version is installed in a different slot. I guess I will have to unmerge (after emerging the older version) the actual version I have installed.
Could somebody tell me if there is something to care about or the correct sequence for doing the glibc downgrade without completely breaking my system ? |
don't downgrade glibc! applications linked against the new one will stop working after downgrade. there is/was some thread about it in the past, even marked sticky, I think... |
Isn't glibc static ?
Edit - Ignore me  |
|
| Back to top |
|
|
KHBuchner
n00b
Joined: 05 Dec 2004
Posts: 1
|
| Posted: Sun Dec 05, 2004 7:19 pm Post subject: x relocation error |
|
|
Hallo,
it seems the problem that is subject here is already known. See
http://www.gentoo.org/proj/en/hardened/hardenedxorg.xml
But I don't understand why this is also happening on my system, i did not set the use flag "hardened". Perhaps someone with a deeper knowledge can enlighten me.
MfG KH |
|
| Back to top |
|
|
hybrid
Apprentice
Joined: 22 Dec 2003
Posts: 224
Location: Lyon, France
|
| Posted: Sun Dec 05, 2004 7:52 pm Post subject: |
|
|
| i just f****d up my system, i am using an other system now, i just applied the patch as suggested, and during the install it exited with an error, now everything is segfaulting |
|
| Back to top |
|
|
vrln
Guru
Joined: 11 Sep 2004
Posts: 534
Location: Finland
|
| Posted: Sun Dec 05, 2004 8:00 pm Post subject: |
|
|
I had this problem too, didn't know about this thread though so I wasn't quite sure what was causing it. Anyway this is how I fixed it:
unmerged both linux26-headers (both were installed for some odd reason) --> emerged linux26-headers (then I only had the newest version)
Then I emerged gcc glibc and xorg-x11 and everything worked again... |
|
| Back to top |
|
|
Sodki
Guru
Joined: 11 Aug 2003
Posts: 419
Location: London, U.K. & Lisbon, Portugal
|
| Posted: Sun Dec 05, 2004 8:37 pm Post subject: |
|
|
i've solved this problem by installing an earlier version on glibc on a new slot (without unmerging the most recent one) and rebooting.
again, don't unmerge the most recent version of glibc, i screwed up my sistem real bad when i did that. fortunately, i was able to recover it. |
|
| Back to top |
|
|
mallchin
l33t
Joined: 21 Jan 2003
Posts: 655
Location: United Kingdom
|
| Posted: Sun Dec 05, 2004 10:45 pm Post subject: |
|
|
| j79zlr wrote: | I am having the same problems as everyone else, but my net access has also died. I can ping within my local network, but any web address fails. I believe this is bad since I wont be able to emerge sync when this gets fixed, just wondering if anyone else has this problem?
btw, I reemeged glibc after installing the new 2.6.8.1-r1 headers, but didn't have any problems. Then I reemerged with the userlocales flag and this began. |
I got this too. Check you've updated all config files, in particular /etc/init.d/net.eth* becoming a symlink instead, and the changes in /etc/conf.d/net -- check your gateway is set properly, most likely the cause of your error (it was mine).
_________________
6700 @ 2.66GHz, 4Gb RAM, 2 x 500Gb, 8800 GTX, PhysX, X-Fi, 24" Widescreen, Tux mascot |
|
| Back to top |
|
|
timbo
Apprentice
Joined: 29 Jul 2002
Posts: 231
Location: New Zealand
|
| Posted: Sun Dec 05, 2004 11:22 pm Post subject: |
|
|
OK, so I prelinked my system so Xorg is working again, once this is all fixed and our systems have settled down again should I unmerge prelink since it has it's own problems?
Regards
Tim
_________________
Linux User: 303160 |
|
| Back to top |
|
|
feardapenguin
Guru
Joined: 23 Jul 2003
Posts: 414
Location: Texas
|
| Posted: Mon Dec 06, 2004 12:21 am Post subject: |
|
|
Still receiving __guard errors here. I've tried downgrading linux26-headers, prelinking, reinstalling gcc with gcc-config-1.3.7-5, even installing an older glibc in a different slot. No luck.
Now my system is still screwed up and I've got 2 versions of glibc. Is this safe and if not, how do I safely get rid of the older one? Is it enough to just emerge -C glibc-<older.version>.
Man, this sucks. |
|
| Back to top |
|
|
j79zlr
Apprentice
Joined: 05 Dec 2004
Posts: 235
Location: Chicago, IL
|
| Posted: Mon Dec 06, 2004 12:25 am Post subject: |
|
|
| mallchin wrote: | | j79zlr wrote: | I am having the same problems as everyone else, but my net access has also died. I can ping within my local network, but any web address fails. I believe this is bad since I wont be able to emerge sync when this gets fixed, just wondering if anyone else has this problem?
btw, I reemeged glibc after installing the new 2.6.8.1-r1 headers, but didn't have any problems. Then I reemerged with the userlocales flag and this began. |
I got this too. Check you've updated all config files, in particular /etc/init.d/net.eth* becoming a symlink instead, and the changes in /etc/conf.d/net -- check your gateway is set properly, most likely the cause of your error (it was mine). |
Thanks for the reply, I am actually bootstrapping a new install as we speak, doesn't seem like there is a solution to the X relocation error as of now, so hopefully a new install won't have it.
I must say this is why I continue to think FreeBSD's ports system is better, atleast ports are tested before they make it into the tree. |
|
| Back to top |
|
|
OldTango
l33t
Joined: 21 Feb 2004
Posts: 718
|
| Posted: Mon Dec 06, 2004 12:25 am Post subject: |
|
|
After some up grades to my system today I also suffer from this problem. Checking everything I could think of reveled nothing. Rebuilt glibc 4 times, and the one time I atempted to rebuilb xorg-x11, it errored and wouldn't build. Which I belive, will still be the case if I attempt it again.
Checking Vrin's option out I did in fact have two both linux26-headers installed, which I am sure is not a good thing. So I followed his path to problem and after all was finished, I still received the error when trying to start X.
Perlinking with prelink -afmR did allow X to start once again, but as this has been stated here already, this is a very ugly workaround to this BUG.
Lets hope that this can be resolved soon, without an ulgy workaround.......................  |
|
| Back to top |
|
|
Sodki
Guru
Joined: 11 Aug 2003
Posts: 419
Location: London, U.K. & Lisbon, Portugal
|
| Posted: Mon Dec 06, 2004 12:35 am Post subject: |
|
|
| feardapenguin wrote: | | Still receiving __guard errors here. I've tried downgrading linux26-headers, prelinking, reinstalling gcc with gcc-config-1.3.7-5, even installing an older glibc in a different slot. No luck. |
did you restart your system after emerging the older glibc? i ask this because my Xorg loaded after a reboot, but not before.
| Quote: | | Now my system is still screwed up and I've got 2 versions of glibc. Is this safe and if not, how do I safely get rid of the older one? Is it enough to just emerge -C glibc-<older.version>. |
i believe it is safe, that's why it's slotted. you can safely emerge -C glibc-<older.version>. |
|
| Back to top |
|
|
blotto
Tux's lil' helper
Joined: 03 Jul 2003
Posts: 116
Location: Whitehaven UK
|
| Posted: Mon Dec 06, 2004 1:03 am Post subject: |
|
|
| my 1st post wrote: | Got the same prob on 2 of my 3 machines both were updated to new headers and glibc recompiled
If I do an | Code: | | emerge -Cp linux26-headers |
Then both machines indicate that
linux26-headers-2.6.8.1 &
linux26-headers-2.6.8.1-r1
need to be unmerged
Ive cleaned out the headers and emerged 2.6.8.1 on one machine and the 2.6.8.1-r1 on the other,
Now emerging glibc on both to see what happens ? |
Still no luck
Both these machines were built using "nptl" and both were using
linux26-headers-2.6.8.1
glibc-2.3.4.20040808-r1
gcc-3.3.4-r1 prior to emerging 2.6.8.1--r1 headers and recompiling glibc.
One of the machines was installed a week ago using these.
This was part of an emerge -uD world and there were only a few packages to update and both were practically identical as follows
sys-kernel/linux26-headers-2.6.8.1-r1
sys-libs/glibc-2.3.4.20040808-r1 (remerge on both systems)
sys-devel/gcc-config-1.3.6-r4
x11-base/opengl-update-1.8.2
sys-libs/db-4.1.25_p1-r4
dev-libs/glib-2.4.7
sys-kernel/gentoo-dev-sources-2.6.9-r9
net-nds/portmap-5b-r9
app-text/docbook-xml-dtd-4.3
sys-devel/bin86-0.16.13
games-arcade/crack-attack-1.1.11-r1
I have tried remerging xorg on both machines and it fails on both machines complaining about | Quote: | undefined reference to --stack_smash_handler
&
undefined reference to --guard |
Havnt tried the prelink thing cause it doesn't sound like a fix or an explanation more a desparate workaround
One of the machines was actually prelinked before and the other was not
Gona hava cuppa an hava think! |
|
| Back to top |
|
|
feardapenguin
Guru
Joined: 23 Jul 2003
Posts: 414
Location: Texas
|
| Posted: Mon Dec 06, 2004 1:16 am Post subject: |
|
|
| Quote: | | i believe it is safe, that's why it's slotted. you can safely emerge -C glibc-<older.version>. |
DO NOT TRY THIS AT HOME! I tried unmerging the old glibc version. The emerge -C blew up before it could even finish. I found myself looking at a dysfunctional command prompt (i.e. no portage, no bash, not even an 'ls' worked). A reboot just resulted in kernel panic and a hole in my wall about the size of a 17" monitor.
I'm restoring from a backup now.
If I had any confidence that I wouldn't run into this again I would consider installing from scratch.
Did I mention that this sucks! |
|
| Back to top |
|
|
NightMonkey
Guru
Joined: 21 Mar 2003
Posts: 356
Location: Philadelphia, PA
|
| Posted: Mon Dec 06, 2004 1:28 am Post subject: Solved for me |
|
|
OK, this worked for me. It was in the bug which is now closed:
| Quote: | ------- Additional Comment #12 From Heinrich Nirschl 2004-12-05 09:31 PST -------
Here is what I did to fix it. I changed the ebuild glibc-2.3.4.20040808-r1.ebuild to apply the glibc-2.3.2-propolice-guard-functions-v3.patch on all architectures and not just for hppa. That is, in function do_ssp_patches() comment out the lines starting with "if" and "fi". After re-emerging glibc all was working again.
It has nothing to do with the new headers, I am using them now.
I guess there was a reason to apply this patch only on hppa. So don't take this as a fix for the problem. It's just a hint. |
Still don't have a solid idea of why this would have affected only one of the two machines upon which I'd rebuilt glibc for nptl support. The evidence does seem clear, however, that it has nothing to do with the linux26-headers or gcc - just glibc. YYMV, however. Good luck! |
|
| Back to top |
|
|
feardapenguin
Guru
Joined: 23 Jul 2003
Posts: 414
Location: Texas
|
| Posted: Mon Dec 06, 2004 1:40 am Post subject: |
|
|
Even the author of that comment admits that it is a workaround, not a fix. Why is the bug closed?
The whole Gentoo experience is leaving a bad taste in my mouth. I want my money back. (Dude, it was free... Oh, right.) |
|
| Back to top |
|
|
blotto
Tux's lil' helper
Joined: 03 Jul 2003
Posts: 116
Location: Whitehaven UK
|
| Posted: Mon Dec 06, 2004 1:49 am Post subject: |
|
|
| Quote: | | Even the author of that comment admits that it is a workaround, not a fix. Why is the bug closed |
I think I tend to agree-
One machine has compiled and used glibc-2.3.4.20040808-r1 for over a month and the other for 9 days until the remerge problem
Whats changed ? |
|
| Back to top |
|
|
feardapenguin
Guru
Joined: 23 Jul 2003
Posts: 414
Location: Texas
|
| Posted: Mon Dec 06, 2004 2:06 am Post subject: |
|
|
I've seen a number of workarounds mentioned in this thread. Some have worked for a few but not for all.
Has anyone else been successful with the suggestion in the bug? I only see two other comments where people said it worked for them. I'd like to know if it works across the board and not just a few isolated cases.
I'm restoring from a 3 week old backup at the moment. Once up I don't plan to update anything related to gcc or base libraries until there is a solution to this. I've already lost my system once. That's enough. |
|
| Back to top |
|
|
element-82
n00b
Joined: 23 Nov 2004
Posts: 6
|
| Posted: Mon Dec 06, 2004 2:09 am Post subject: |
|
|
Commenting out a few "ifs" in a patch, which worked fine previously without the workaround, is not a fix. The only thing I didn't try was a reboot, I'll try anything once. Otherwise, I'll have to reinstall.
I shall change my signature to "Never update glibc in gentoo, that is akin to standing in the silly spot".
Pb
| blotto wrote: | | Quote: | | Even the author of that comment admits that it is a workaround, not a fix. Why is the bug closed |
I think I tend to agree-
One machine has compiled and used glibc-2.3.4.20040808-r1 for over a month and the other for 9 days until the remerge problem
Whats changed ? |
|
|
| Back to top |
|
|
blotto
Tux's lil' helper
Joined: 03 Jul 2003
Posts: 116
Location: Whitehaven UK
|
| Posted: Mon Dec 06, 2004 2:24 am Post subject: |
|
|
Also theres my 3rd (main) box thats been also running glibc-2.3.4.20040808-r1 since 15 Nov with 2.6.8.1 headers no problem and like the other two is using xorg-x11-6.8.0-r3
Guess what emerge -upD world is tempting me with --NOT |
|
| Back to top |
|
|
feardapenguin
Guru
Joined: 23 Jul 2003
Posts: 414
Location: Texas
|
| Posted: Mon Dec 06, 2004 4:32 am Post subject: |
|
|
I just restored my system from a "cloned" backup dated Nov 14th. The clone process creates a mirror image of the system onto another drive or partition.
The newly re-cloned system contains the following configuration:
| Code: | gcc-3.4.3
gcc-config-1.3.7-r1
glibc-2.3.4.20041102
linux26-headers-2.6.8.1-r1
|
Note that this system has NOT been updated (i.e. emerge sync'd) since 11/14. After re-cloning I immediately reemerged the same versions of gcc-3.4.3 and glibc-2.3.4.20041102 with no problem using the ebuilds from 11/14. No surprise there.
It is interesting to note, however, that the "do_ssp_patches" function is intact in the old glibc ebuild (see below) AND RAN UNCOMMENTED WITH NO PROBLEM.
| Code: | do_ssp_patches() {
# To circumvent problems with propolice __guard and
# __guard_setup__stack_smash_handler
#
# http://www.gentoo.org/proj/en/hardened/etdyn-ssp.xml
if [ "${ARCH}" != "hppa" ] && [ "${ARCH}" != "hppa64" ]; then
epatch ${FILESDIR}/2.3.3/glibc-2.3.2-propolice-guard-functions-v3.patch
cp ${FILESDIR}/2.3.3/ssp.c ${S}/sysdeps/unix/sysv/linux || \
die "failed to copy ssp.c to ${S}/sysdeps/unix/sysv/linux/"
fi
# patch this regardless of architecture, although it's ssp-related
epatch ${FILESDIR}/2.3.3/glibc-2.3.3-frandom-detect.patch
} |
This, if nothing else, convinces me that real problem is not with this patch alone. Something else must have borked it.
If will be interesting to emerge sync and see what changes. Since I've still got the cloned system I should be able to diff around and compare the two system states. |
|
| Back to top |
|
|
John-Boy
Guru
Joined: 23 Jun 2004
Posts: 442
Location: Desperately seeking moksha in all the wrong places
|
| Posted: Mon Dec 06, 2004 5:57 am Post subject: |
|
|
We can say that it's something with glibc then, not the headers. Is it worth filing another bug report, does anybody think ? I know there are two, last time I check both were marked as closed.
I'm not touching anything this side until the dust settles a bit on this. |
|
| Back to top |
|
|
jarealist
Apprentice
Joined: 07 Oct 2002
Posts: 228
|
| Posted: Mon Dec 06, 2004 6:07 am Post subject: |
|
|
Just saw this in the glibc changelog after the latest sync:
"05 Dec 2004; <solar@gentoo.org> glibc-2.3.4.20041102.ebuild:
- q/a - missed glibc-2.3.4.20041102 in last fix
05 Dec 2004; <solar@gentoo.org> glibc-2.3.4.20040808-r1.ebuild:
- q/a fix - hppa is the only arch with an upwords growing stack and thus can
not use ssp vs being the only arch that can use ssp"
Maybe this is supposed to fix the problem, but I'm still going to wait a little. |
|
| Back to top |
|
|
unlocked
n00b
Joined: 24 Oct 2004
Posts: 53
Location: Austin, Tx.
|
| Posted: Mon Dec 06, 2004 6:41 am Post subject: |
|
|
I was going through my logs and saw that on Nov 26 it applies the hppa guard patch, but on Dec 5 it didn't so I am applying it and rebuilding glibc then hopfully emerge -e system will fix it. That will teach me to keep tinkering with cflags. So I guess it was just a .ebuild update from a --sync that fubared it.
There could be some other ebuild issues. I just want to get system back to building emerge -euD world again.
_________________
abit av8 (939), 3000+, 1gig ram, ti4400, dc10+, 120gig sata, audigy2. |
|
| Back to top |
|
|
|
Desktop Environments
