Dear Sebastian,
In message E70AF999396FDF4EAE40E195B847096109F5D6FD@SRVEXCH-2K10.CADCON.INTERN you wrote:
We use a custom board configuration for our custom hardware. I can send parts of it, if needed.
Thanks, but no. Parts are of no use to us. Basicly you have to options: have your code merged into mainline (to do that you have to submit a complete set of patches here on the mailing list), or maintain your out-of-tree port yourself (in which case David Woodhouse's statement applies, see http://article.gmane.org/gmane.linux.kernel.embedded/3534).
We are currently using v2012.10.
This works fine for the mainline code.
The toolchain is OSELAS.2012.12.0 (http://www.pengutronix.de/oselas/toolchain).
So eother it's an issue with your port, or with your toolchain.
With toolchain OSELAS.2011.11.2 everything worked fine. We didn't change anything else.
Then this smells like a tool chain issue. You might contact Pengutronix support for help with their tool chain.
So I suppose this might be a mainline problem. It seems like the mentioned variables don't get set correctly as before.
Well, no such problems are observed with mainline code and - for example - ELDK 5.5 and 5.4-M3-rc1. What makes you think it's a mainline issue? Can you prove such problems with mainline code, and with a known to be working tool chain?
Best regards,
Wolfgang Denk