Hide Forgot
Created attachment 487155 [details] yum.log after "yum upgrade" hang Description of problem: "yum upgrade" hangs on glibc-common on fresh Fedora 15 alpha install Version-Release number of selected component (if applicable): yum-3.2.29-4.fc15.noarch How reproducible: 100% Steps to Reproduce: 1. Do a fresh install of Fedora 15 x86_64 alpha (install DVD) 2. yum upgrade Actual results: Yum hangs when updating glibc-common-2.13.90-7.x86_64. Expected results: Yum completes the upgrade successfully. Additional info: I installed Fedora 15 alpha the day it was released. Mostly daily "yum upgrade"s worked fine (sometimes with --skip-broken). The yum upgrade on March 22 (yesterday) hung at glibc-common-2.13.90-7.x86_64. Same thing today. I have repeated this a number of times including twice from a fresh install running yum from the console (ie, no Gnome). yum upgrade with "-e 10 -d 10" shows nothing interesting. There are "Adding Package" messages up til "Running Transaction" after which there are no messages other than the normal Updating/Installing messages. yum.log just contains one entry for each of the packages updated or installed to that point. There is no entry for glibc-common. After 1 hour, I ctrl-C'ed yum. It says: Non-fatal POSTIN scriptlet failure in rpm package glibc-common-2.13.90-7.x86_64 Related "yum check" output is (will attach the full output): glibc-common-2.13.90-7.x86_64 is a duplicate with glibc-common-2.13.90-3.x86_64 glibc-common-2.13.90-7.x86_64 has missing requires of glibc = ('0', '2.13.90', '7')
Created attachment 487157 [details] Full "yum check" output
This is a glibc problem, and apparently quite hardware specific: it happens 100% reliably on my system (Intel Core i5-2500K) but I can't reproduce it in a virtual box. On update to glibc-common-2.13.90-7.x86_64, /usr/sbin/build-locale-archive gets stuck in a busy loop, and if you kill that process this starts happening to many many other commands too, resulting in a very broken system. Here's what gdb says on the stuck /usr/sbin/build-locale-archive process during update: Attaching to process 5395 Reading symbols from /usr/sbin/build-locale-archive...Reading symbols from /usr/lib/debug/usr/sbin/build-locale-archive.debug...done. done. 0x000000000044293a in intel_check_word (name=194, value=29360191, has_level_2=0x7fff4622433f, no_level_2_or_3=0x7fff4622433e) at ../sysdeps/x86_64/cacheinfo.c:196 196 asm volatile ("xchgl %%ebx, %1; cpuid; xchgl %%ebx, %1" (gdb) bt #0 0x000000000044293a in intel_check_word (name=194, value=29360191, has_level_2=0x7fff4622433f, no_level_2_or_3=0x7fff4622433e) at ../sysdeps/x86_64/cacheinfo.c:196 #1 0x0000000000442b21 in handle_intel (name=194, maxidx=<optimized out>) at ../sysdeps/x86_64/cacheinfo.c:338 #2 0x00000000004011b1 in init_cacheinfo () at ../sysdeps/x86_64/cacheinfo.c:570 #3 0x0000000000408767 in __libc_csu_init (argc=1, argv=0x7fff46224498, envp=0x7fff462244a8) at elf-init.c:141 #4 0x0000000000407ff8 in __libc_start_main (main=0x400cda <main>, argc=1, ubp_av=0x7fff46224498, init=0x4086f0 <__libc_csu_init>, fini=0x408780 <__libc_csu_fini>, rtld_fini=0, stack_end=0x7fff46224488) at libc-start.c:185 #5 0x0000000000401401 in _start () (gdb) No idea what it's really doing there, but quite obviously something Intel-specific.
The endless loop has been noticed (and supposedly also fixed) at upstream too: http://sourceware.org/bugzilla/show_bug.cgi?id=12587
glibc-2.13.90-8 has been submitted as an update for Fedora 15. https://admin.fedoraproject.org/updates/glibc-2.13.90-8
My failure is also with Intel Core i5-2500K (on Intel DH67GD motherboard). I couldn't imagine hardware being part of this. I should have added that to the initial report.
A "me too" on Intel DH67GD board and Core i5 2500K, exactly the same symptoms, only resolvable by booting from CD and "rpm -r" downgrading.
glibc-2.13.90-8 has been pushed to the Fedora 15 stable repository. If problems still persist, please make note of it in this bug report.
"yum upgrade" now completes for me with glibc-2.13.90-8