rohc team mailing list archive

[Nick Wernicke <nick.wernicke@xxxxxxxxx>]

Regarding the logging issue when not specifying --basedev: There is no
"underlying interface necessary" message in any logs in /var/log/. This is
a Debian 7 VM, I suspect it has something to do with that distribution. It
is not a big deal once I opened the code and saw the proper usage.

I am able to get two VM's to talk to one another and stream video through
the tunnel, one as client, the other as server.

I am not able to get my embedded system client to connect to the server.
The log says:
"TLS handshake succeeded
Certificate cannot be verified (status 1026)"
I dont think there is any problem in iprohc. I think the gnutls_handshake
function in gnutls handles the tls hanshake and certificate verification,
and it is returning unsuccessful. I don't understand what the server gnutls
doesnt like about my certificate.

Certificate was generated and signed on the server VM. I tried the cert
that worked on the client VM, that also did not work on the embedded board.
Could it be a gnutls version issue? Or could there be some additional ssl
related dependency that is not on the embedded board? Or is there something
else it could be?
On Sep 20, 2014 7:36 AM, "Didier Barvaux" <didier@xxxxxxxxxxx> wrote:

> Nick,
>
> My answers below.
>
>
> > At some point, my application will require a cross compiled (ARM)
> > client on embedded linux connected via a low bandwidth link to a
> > server running on Windows 7. I'd like to use iprohc since it is
> > already written, but I am concerned about the dependencies on the
> > minimalist embedded system and windows.
> >
> > --->Is there a way to cross-compile all the dependencies into a single
> > client or server executable for different kinds of deployments? I am
> > not very experienced with the details of cmake...
>
> For the minimalist embedded Linux system, it is possible to build a
> static binary of the IP/ROHC client or server.
>
> For Windows, well... the IP/ROHC code base doesn't support Windows at
> all.
>
>
> > I generated an eclipse project and opened the code with $ cmake -G
> > "Eclipse CDT4 - Unix Makefiles". I can build it and debug it in
> > eclipse with GDB. The program terminates in server.c after line 450:
> > if(strcmp(server_opts.basedev, "") == 0)
> > {
> > trace(LOG_ERR, "underlying interface name is mandatory");
> > goto free_client_contexts;
> > }
> >
> > --->I suspect I did not configure my application correctly. There is
> > probably an argument needed. But which argument? Also, I have one
> > other question: the generated Eclipse project shows a lot of errors,
> > all of them are unresolved symbols such as LOG_ERR, IFNAMSIZ, or
> > LOG_DEBUG. They should be #defined somewhere or cmake -DVAR=value,
> > but I don't see those symbols defined anywhere in the project. It
> > could be that I need to add an include folder into the project that
> > the project builder didn't add. Where are all the definitions?
>
> Yes, as you reported in your other email, you should use the --basedev
> (or -b) option to specify the base network device for the tunnel.
>
> What looks strange to me is that you said that didn't get the
> "underlying interface name is mandatory" message in the logs. Could you
> please check the logs again?
>
> Note that next versions of IP/ROHC (not released yet) have a better
> warning message that reads "wrong usage: underlying interface name is
> mandatory, use the --basedev or -b option to specify it".
>
> Regards,
> Didier
>