[Yum] redirected yum output strangely destroyed by null characters on dist-install and upgrade

[Alexander 'Leo' Bergolth]

Hi!

Maybe this is the wrong list to ask, since this might be some
kernel/libc-issue but it happened during a dist-upgrade using yum... ;-)

When running a distribution upgrade with stdout and stderr redirected to
a file, this file gets truncated during installation of the new kernel.
The truncated file contains a large block of zeroes at the start.
It looks like truncation happens during mkinitrd is running.

I've seen this behavior on one box that was upgraded from F7 to F10
while running the F7 kernel and on one box that was installed using
anaconda started from a F10 Live CD. (This time, the install.log,
written by anaconda was truncated.) A second run of the same install
finished without showing that problem.

I captured the output of the whole F7->F10 upgrade process by viewing
the output using "tail -f" on another box and capturing it's output with
screen. The mangled upgrade-log and the capture of "tail -f" can be
viewed at
  http://leo.kloburg.at/tmp/rpm-output/upgrade.txt
and
  http://leo.kloburg.at/tmp/rpm-output/upgrade-tail-f.txt

(Look for "tail: upgrade.log: file truncated" in line 1399 of
upgrade-tail-f.txt. Use truncate-lines if viewing with emacs!)

I have no idea how such bad things can happen unless some kernel issue
is involved, since the output was redirected using ">&", so stdout and
stderr of yum should be directly connected to that file.

Any ideas?

Cheers,
--leo
-- 
e-mail   ::: Leo.Bergolth (at) wu-wien.ac.at
fax      ::: +43-1-31336-906050
location ::: Computer Center | Vienna University of Economics | Austria