[12:07] <bat0> hello
[08:04] <yong> heh
[04:55] <mark> cjwatson: I had our on-site tech investigate that "man-db" problem
[04:55] <mark> which isn't man-db but most often occurs during that postinst
[04:55] <mark> it seems to be a deadlock
[04:55] <mark> perl seems to try to read from a dpkg pipe
[04:55] <mark> and dpkg is wait4() for perl
[04:56] <mark> at least that's what my on-site tech tells me, a little hard to get more info
[05:23] <mark> I should just always run it in strace, then the deadlock does not happen due to other buffering/context switching ;)
[07:51] <cjwatson> perl would be debconf
[07:52] <cjwatson> so again, it's probably a failure to close some file descriptor, possibly racily
[08:25] <mark> not the commonly made mistake of writing and reading from a child processes's pipes at the same time? :)
[08:25] <cjwatson> I sincerely hope not
[08:25] <cjwatson> no, more usual with debconf is spawning something from a debconf confmodule that daemonises but forgets to close the debconf fds
[08:26] <cjwatson> man-db does background itself under certain circumstances, but I thought that (a) it didn't do that under the noninteractive frontend to avoid pretty much this problem and (b) it disconnected properly anyway ...
[08:28] <mark> ok
[08:28] <mark> yeah but it's not just man-db
[08:28] <mark> for example, today sometimes it got stuck on openssh
[08:28] <mark> and occasionally on other stuff
[08:28] <mark> just most often man-db
[08:29] <mark> today we reinstalled 20 servers and only *one* went through without problems
[08:29] <cjwatson> your libc isn't spawning some weird NSS thing, is it?
[08:43] <mark> ehm
[08:43] <mark> our libc?
[08:43] <mark> it's just the ubuntu installer :)
[08:43] <mark> not doing anything special in that regard
[10:30] <BadMike> hello can anyone help me installing software