Possible regression from timeout commit for resolv.conf

Kevin Day thekevinday at gmail.com
Tue Apr 6 22:02:04 UTC 2010


On 4/6/10, Natanael Copa <natanael.copa at gmail.com> wrote:
> On Fri, Apr 2, 2010 at 12:21 AM, Kevin Day <thekevinday at gmail.com> wrote:
>  > I am using the NPTL branch on x86, which works perfectly fine (in
>  > fact, better than any other stable release since 0.9.28)
>  > The commit number is: cca45baf8353d1e338d232f5bdb2d1d6b357f1da
>  > (http://git.uclibc.org/uClibc/commit/?h=nptl&id=cca45baf8353d1e338d232f5bdb2d1d6b357f1da)
>  >
>  > With the addition of the commit, the program claws-mail fails to
>  > respond to SMTP servers and throws an annoyingly useless error.
>  > At first glance I assumed that this to be a claws-mail problem.
>  >
>  > Somehow adding support for timeout and attempts in resolv.conf breaks
>  > claws-mail.
>  >
>  > When I revert those changes in mentioned commit, recompile uClibc, and
>  > run claws-mail again, it works.
>
>
> FWIW, I have issues with claws-mail with imaps on uclibc-0.9.30.x. SSL
>  handshake error. I think its gnutls causing that. sylpheed with
>  openssl works just fine.
>
>  --
>
> Natanael Copa
>

The SSL portions work if I use an IP address instead of a domain name.
I get a certificate, accept, and continue.
Mails in both directions work.

So if this were SSL, then it is an SSL+resolve issue and not just a
resolve issue.
I am also wondering if it is libetpan, which claws-mail includes and uses.

FYI: I am using the following NPTL commits because they work rather well:
older version: 784e157fadbf1622d381e7dc9e82d2510a690776
newer version: f7eba78a7464c7b31326abf92dab254608835028

More recent ones prior to the resync with 0.9.3* do not compile.
I have yet to get around to trying the NPTL after the resync to 0.9.3*.

-- 
Kevin Day


More information about the uClibc mailing list