TW glibc 2.33 "CPU ISA level is lower than required"
Hi, I just tried to update a 32bit TW system from snapshot 20210210 to 20210218. After the update of glibc the further update fails and no new programs can be started: ( 69/1833) Installing: glibc-2.33-1.1.i686 ...............................................................................................................................................[done] Additional rpm output: /lib/libc.so.6: CPU ISA level is lower than required ( 70/1833) Installing: translation-update-de-15.1-2.3.noarch ............................................................................................................................[error] Installation of translation-update-de-15.1-2.3.noarch failed: Error: Subprocess failed. Error: RPM failed: /lib/libc.so.6: CPU ISA level is lower than required Abort, retry, ignore? [a/r/i] (a): Is this the end of Tumbleweed on x86_32 hardware, or is this a bug? Kind regards, Dieter
Glibc maintainers need to add libc_cv_include_x86_isa_level=no to configure
because -march is being set on i686
--
Callum Farmer
gmbr3@opensuse.org
openSUSE/GitHub/GitLab - gmbr3
On Sat, 20 Feb 2021, 08:27 dieter,
Hi,
I just tried to update a 32bit TW system from snapshot 20210210 to 20210218.
After the update of glibc the further update fails and no new programs can be started:
( 69/1833) Installing: glibc-2.33-1.1.i686
...............................................................................................................................................[done] Additional rpm output: /lib/libc.so.6: CPU ISA level is lower than required
( 70/1833) Installing: translation-update-de-15.1-2.3.noarch
............................................................................................................................[error] Installation of translation-update-de-15.1-2.3.noarch failed: Error: Subprocess failed. Error: RPM failed: /lib/libc.so.6: CPU ISA level is lower than required
Abort, retry, ignore? [a/r/i] (a):
Is this the end of Tumbleweed on x86_32 hardware, or is this a bug?
Kind regards, Dieter
On Saturday 2021-02-20 11:58, Callum Farmer wrote:
Glibc maintainers need to add libc_cv_include_x86_isa_level=no to configure because -march is being set on i686
No, it's not because -march is used. glibc always activates it if binutils has the feature. I now see that glibc.i586 is also polluted, so installing that alternative rpm won't help dieter. Great way to bork all i586 systems >:-(
Specifying march to glibc 2.33 causes it to infer too high an ISA level
Seen at Gentoo and Arch Linux already. Same patch needs doing here.
--
Callum Farmer
gmbr3@opensuse.org
openSUSE/GitHub/GitLab - gmbr3
On Sat, 20 Feb 2021, 12:56 Jan Engelhardt,
On Saturday 2021-02-20 11:58, Callum Farmer wrote:
Glibc maintainers need to add libc_cv_include_x86_isa_level=no to configure because -march is being set on i686
No, it's not because -march is used. glibc always activates it if binutils has the feature. I now see that glibc.i586 is also polluted, so installing that alternative rpm won't help dieter. Great way to bork all i586 systems >:-(
On Saturday 2021-02-20 14:06, Callum Farmer wrote:
Specifying march to glibc 2.33 causes it to infer too high an ISA level
Not what I see. https://sourceware.org/bugzilla/show_bug.cgi?id=27446
On Saturday 2021-02-20 09:27, dieter wrote:
I just tried to update a 32bit TW system from snapshot 20210210 to 20210218.
After the update of glibc the further update fails and no new programs can be started:
( 69/1833) Installing: glibc-2.33-1.1.i686 ...............................................................................................................................................[done] Additional rpm output: /lib/libc.so.6: CPU ISA level is lower than required
I found it is a problem in the generation of glibc.i686. (The file crt1.o encodes a .note.gnu.property section with essentially "requires x86_64" and so, is borked.) You can workaround it by booting a rescue system and installing glibc-2.33-1.1.i586.rpm instead.
On Sat, 20 Feb 2021 12:01:32 +0100 (CET) Jan Engelhardt wrote:
( 69/1833) Installing: glibc-2.33-1.1.i686 ...............................................................................................................................................[done] Additional rpm output: /lib/libc.so.6: CPU ISA level is lower than required
I found it is a problem in the generation of glibc.i686. (The file crt1.o encodes a .note.gnu.property section with essentially "requires x86_64" and so, is borked.)
You can workaround it by booting a rescue system and installing glibc-2.33-1.1.i586.rpm instead.
thanks for the hint Jan, I will try it. Kind regards, Dieter
On Sat, 20 Feb 2021 12:39:58 +0100 dieter wrote:
You can workaround it by booting a rescue system and installing glibc-2.33-1.1.i586.rpm instead.
thanks for the hint Jan, I will try it. I tried it.
I downloaded openSUSE-Tumbleweed-NET-i586-Snapshot20210218-Media.iso because of the i586 (the rescue image would be openSUSE-Tumbleweed-Rescue-CD-i686-Snapshot20210218-Media.iso so I thought I probably do not need to try it). But there happens a kernel panic very early during boot. This kernel panic looks very similar in my VM I am using to test prior to updating the real hardware when I try to boot the netinstall iso on either of them. My test VM has the CPU set to pentium3 or athlon, the bare metal has a Athlon-XP. Kind regards, Dieter
On Saturday 2021-02-20 17:11, dieter wrote:
On Sat, 20 Feb 2021 12:39:58 +0100 dieter wrote:
You can workaround it by booting a rescue system and installing glibc-2.33-1.1.i586.rpm instead.
thanks for the hint Jan, I will try it. I tried it.
I downloaded openSUSE-Tumbleweed-NET-i586-Snapshot20210218-Media.iso because of the i586 (the rescue image would be openSUSE-Tumbleweed-Rescue-CD-i686-Snapshot20210218-Media.iso so I thought I probably do not need to try it).
But there happens a kernel panic very early during boot.
I see it too - the reason is, once again, the problem loading libc.so.6. (Since there is no other program that could execute, the kernel does not know what to do and just stops with a panic.) There is not much that's worthwhile doing other than waiting for glibc to be fixed and TW.i586 to have rebuilt with a new ISO.
On Saturday 2021-02-20 21:03, Jan Engelhardt wrote:
I downloaded openSUSE-Tumbleweed-NET-i586-Snapshot20210218-Media.iso because of the i586 (the rescue image would be openSUSE-Tumbleweed-Rescue-CD-i686-Snapshot20210218-Media.iso so I thought I probably do not need to try it).
But there happens a kernel panic very early during boot.
I see it too - the reason is, once again, the problem loading libc.so.6. (Since there is no other program that could execute, the kernel does not know what to do and just stops with a panic.)
There is not much that's worthwhile doing other than waiting for glibc to be fixed and TW.i586 to have rebuilt with a new ISO.
http://download.opensuse.org/history/20210210/tumbleweed/repo/oss/boot/i386/ one can TFTP-boot "linux" and "initrd" and then point linuxrc to use the same location (20210210) to get into a rescue system. But that does not help much, since all current Tumbleweed i586 binaries have the problematic x86_64 require, so my early suggestion to install glibc-2.33-1.1.i586.rpm wouldn't help (not sure which copy of glibc-without-ISA I had looked at).
On Sat, 20 Feb 2021 21:50:42 +0100 (CET) Jan Engelhardt wrote:
There is not much that's worthwhile doing other than waiting for glibc to be fixed and TW.i586 to have rebuilt with a new ISO.
http://download.opensuse.org/history/20210210/tumbleweed/repo/oss/boot/i386/ one can TFTP-boot "linux" and "initrd" and then point linuxrc to use the same location (20210210) to get into a rescue system.
But that does not help much, since all current Tumbleweed i586 binaries have the problematic x86_64 require, so my early suggestion to install glibc-2.33-1.1.i586.rpm wouldn't help (not sure which copy of glibc-without-ISA I had looked at).
Thanks for the suggestions. Currently I have no problem. I encountered this while updating a VM. I can just fall back to the previous state by discarding the image-snapshot which contains the problematic glibc update and wait with the next update until a fix is available. Kind regards, Dieter
participants (3)
-
Callum Farmer -
dieter -
Jan Engelhardt