Something in the latest bunch of updates is triggering an rpm bug, or
rather, a librepo bug.
I ran a dnf upgrade on a bunch of machines. The first one segfaulted in the
cleanup phase. While I was pondering what to do about it, maybe a cosmic ray
hit an unfortunate capacitor in one of the RAM sticks perhaps, another one
splurted this out:
Transaction Summary
================================================================================
Install 5 Packages
Upgrade 159 Packages
Remove 4 Packages
Total download size: 514 M
Is this ok [y/N]: y
Downloading Packages:
corrupted double-linked list
Aborted (core dumped)
A core dump showed this:
Stack trace of thread 71232:
#0 0x00007f401e6bf9e5 raise (libc.so.6 + 0x3c9e5)
#1 0x00007f401e6a8895 abort (libc.so.6 + 0x25895)
#2 0x00007f401e703857 __libc_message (libc.so.6 + 0x80857)
#3 0x00007f401e70ad7c malloc_printerr (libc.so.6 + 0x87d7c)
#4 0x00007f401e70babc unlink_chunk.constprop.0 (libc.so.6 +
0x88abc)
#5 0x00007f401e70e3aa _int_malloc (libc.so.6 + 0x8b3aa)
#6 0x00007f401e710235 __libc_calloc (libc.so.6 + 0x8d235)
#7 0x00007f4016b45f85 lr_malloc0 (librepo.so.0 + 0x1ef85)
#8 0x00007f4016b37e55 lr_downloadtarget_new (librepo.so.0 +
0x10e55)
#9 0x00007f4016b40842 lr_download_packages (librepo.so.0 +
0x19842)
#10 0x00007f4016ca32ea
_ZN6libdnf13PackageTarget16downloadPackagesERSt6vectorIPS0_SaIS2_EEb
(libdnf.so.2 + 0x14a2ea)
#11 0x00007f40151ed986 _wrap_PackageTarget_downloadPackages
(_repo.so + 0x1e986)
This strongly suggests a bone-fide bug. A rerun of 'dnf upgrade' succeeded
without any fanfare.
I'm fully updated out, but perhaps someone who has a bunch of updates to
install could run dnf upgrade under valgrind, and see what shakes out.
Show replies by date