Red Hat Bugzilla – Bug 185930
anaconda --rootpath does not restrict itself to rootpath
Last modified: 2007-11-30 17:11:27 EST
Description of problem:
Under the assumption that anaconda --rootpath should only be modifying the
contents of rootpath (not other files or state on the host system), there are at
least 3 bugs currently breaking this behavior.
- reconfigures host network
- reconfigures host timezone
- leaves a couple symlinks in /etc.
See kadischi bug 185830 for complete details
Version-Release number of selected component (if applicable):
Steps to Reproduce:
I'm pretty sure I've at least tracked down the timezone part of this bug-
grepping for localtime in /usr/lib/anaconda/*.py, I see that /etc/localtime gets
written in two places.
in timezone.py, the target installed system's localtime (I.E. instPath +
"/localtime) gets set
iutil.copyFile(fromFile, instPath + "/etc/localtime")
That is as it should be. The problem however is in packages.py in
setupTimezone, where this happens (note lack of instPath)
There is already a fallthrough in setupTimezone for testing, i.e.
# dont do this in test mode!
The problem is that this is definately a bad thing to do in --rootpath mode when
rootpath mode is used by kadischi to generate a livecd. I'm guessing this is a
bad thing to do in rootpath mode in general, but I don't know the anaconda
userbase well enough to say for certain.
I suppose it might be too much to ask for a reference to a design spec which
would define what exactly rootpath should and should not be doing and/or used
for? I'm just wondering how closely rootpath maps to what kadischi is using it
for, and if another option is necessary.
Fixed all of the cases caught here. If you notice anything else, file them and
I'll take care of them.