Michael Braun wrote:
> Hi,
>
> I've had a problem with both licq and kopete under Mandrake (cooker). I
> can start the software but did not get any network connections with an
> icq server. Am I the only guy who has this problem? Should it be a
> problem with the newest libqt3 ? I use KDE3.1(RC5).
I am having the same problem with CVS builds of KDE, and I believe that's a
glibc problem.. As a really evil workaround you can put in the IP addresses
for the ICQ servers into /etc/hosts, that should make it "work".(There is
an another workaround/evil-hack involving patching Qt, but I don't think
you want to rebuild it yourself)
(I've submitted the info on the apparent bug to glibc-bug-new, but I can't
really use the glibcbug script as I don't have an MTA, and I am afraid it
didn't seem to show up). Basically, when the following is built with
-pthread:
#include <netinet/in.h>
#include <arpa/nameser.h>
#include <resolv.h>
#include <netdb.h>
#include <stdio.h>
int main()
{
printf("(Initial) id = %d options = %d\n", _res.id, _res.options);
res_init();
printf("(After res_init) id = %d options = %d\n", _res.id,
_res.options);
printf("Inited = %d DNS count=%d\n", _res.options & RES_INIT,
_res.nscount);
gethostbyname("localhost");
printf("(After gethostbyname) id = %d options = %d\n", _res.id,
_res.options);
printf("Inited = %d DNS count=%d\n", _res.options & RES_INIT,
_res.nscount);
}
The result is:
(Initial) id = 0 options = 0
(After res_init) id = 0 options = 0
Inited = 0 DNS count=0
(After gethostbyname) id = 27443 options = 705
Inited = 1 DNS count=1
Which basically means that, when you ask glibc to initialize the resolver
library, it doesn't. Qt (QDns) uses that to get a list of DNS servers. And
as a result, you end up running an app that can't possibly lookup the IP
addresses for the IM servers.
This worked fine in earlier/original glibc-2.3.1 packages; it has also been
confirmed/reproduced by people using Debian Sid glibc-2.3.1-8 packages; and
not glibc-2.3.1-5; the difference there as in cooker seems primarily in
updating along the 2.3.x branch -- so that may be a regression there...
(Hoping that Gwenole Beauchesne notices this),
-Maksim
