On Mon, 11 Aug 2008 somebody known as Jurij Smakov wrote:
Hi,
The line invoking ucf in the postinst looks like that:
test -f /usr/bin/ucf && ucf ${CONFFILE_NEW} ${CONFFILE} </dev/tty
Now, I don't know why you would want to get it read from /dev/tty, as
far as I can tell, ucf does not read anything from stdin, so this is
pretty redundant. However, once I removed this redirect, it was still
hanging. Eventually I traced the problem to the fact that ucf uses
debconf, I can't fully understand what's happening there, but it
turned out that moving db_stop, placing it *after* ucf invocation
fixes the problem. I guess that once we stop debconf in the script
(with db_stop) and then ucf tries to start it up again, something goes
wrong with file descriptors. The candidate patch which fixes the issue
is attached, however I'm reluctant to declare it a proper fix without
understanding what's really going on, hence I'm not tagging it.
Thanks for investigating this issue. Although I'm maintainer of this
package, my defconf and ucf understanding is rather limited (I'm upstream
author, so I eneded up as maintainer of bunch of other packages: for
Gentoo, OpenWRT etc).
Whould it be possible to get rid of the ucf completely?
dibbler-client.postinst could generate new version (and rename
old one) without any ucf invocation. As I understand it, the only
flaw would be that this backed up file would not be marked as config. file
for dibbler-client.
Regards,
--
Tomasz Mrugalski, | "We all know Linux is great...it does |
thomson(at)klub(dot)com(dot)pl | infinite loops in 5 seconds." |
| Linus Torvalds |
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
