From: Steve Kersley <steve.kersley@ke...> - 2002-09-03 16:55:52
> Try building your UML kernel with debugging support, so you can start it up
> with the 'debug' flag. You should then be able to start to compile MySQL as
> usual, and when it gets 'stuck', issue a 'Ctrl-C' on gdb and do a backtrace
> to see what UML has been up to. If it's using CPU time, then it'll be stuck
> in a loop somewhere, one would imagine.
OK, I'll give that a go and see what it turns up. I forgot to mention that
uml is running as a non-root user on the host - I don't imagine that should
cause any problems, but you never know.
Also, a bit of nosing around has shown that the file it gets stuck on (and
indeed the function it claims to have been working on when I eventually
killed the process) looks like it uses pthreads. Not sure if there are any
From: Jeff Dike <jdike@ka...> - 2002-09-03 17:58:26
> I'm not sure whether this is a problem with gentoo, usermode or mysql
> (or even if I'm just being impatient, but this is the third time I've
> tried to compile with the same problem - on another system, no
Was that other system a UML or a physical box?
> Anyone got any ideas of anything to try?
Can you get some stack traces from UML during this period?
> Guest kernel is 2.4.18-2um.
Could you try a more modern UML and see if it behaves better?
From: David Coulson <david@da...> - 2002-09-03 16:58:14
> OK, I'll give that a go and see what it turns up. I forgot to mention that
> uml is running as a non-root user on the host - I don't imagine that should
> cause any problems, but you never know.
The whole point of UML is so that you don't have to run stuff as root on
the host. If you were running UML as root on the host, we'd have to send
someone 'round to teach you how to do it properly. ;-)
There's no obvious reason why that should make a difference - You
compile MySQL as a user on the host, right?
> Also, a bit of nosing around has shown that the file it gets stuck on (and
> indeed the function it claims to have been working on when I eventually
> killed the process) looks like it uses pthreads. Not sure if there are any
> issues there.
Unless Gentoo has something messed up with it's compiler, it shouldn't
matter. MySQL with pthreads runs happily under UML here, although it was
from a Debian package rather than compiled.
David Coulson http://davidcoulson.net/