I'm using pidgin on my laptop so often I use it on different networks. When I my network setting automagically changes in function of the net I use (I use fplugd + suspend2-sources, but a manual change has the same effects) pidgin stops working. If I try to disconnect and then reconnect pidpig it reports errors like this: "Could not connect to authentication server [...] Temporary failure in name resolution" but the others applications like thunderbird, firefox, mail-notification work great in the same conditions. Speaking with a pidgin dev (on their irc channel) he told me that pidgin simply calls gethostbyname() and there's no caching of DNS server data, so he thinks it would be a bug in my (gentoo) libc's DNS resolution library
This bug is strange, do we have anything special in our libc resolver configuration?
Why would only pidgin be affected though? Is the reporter running any local dns or caching services like these? dnsmasq, bind, djbdns, nscd
(In reply to comment #2) > Is the reporter running any local dns or caching services like these? > dnsmasq, bind, djbdns, nscd No, I haven't any of these packages installed.
nscd comes with glibc, which you do have installed. When the switch happens, is /etc/resolv.conf still valid compared to the previous one?
this is a "feature" of the glibc resolver which is "solved" by using nscd see: http://sources.redhat.com/bugzilla/show_bug.cgi?id=984 http://sources.redhat.com/bugzilla/show_bug.cgi?id=3675
SuSE has a patch for this which i'll merge since i agree this is a stupid feature
So you don't want the dhcp clients to do ncsd -i hosts if they cannot do res_init then?
i dont understand your question
