broken foregroundlock in codegentest.cpp

[floris]

On Thu, 2010-12-30 at 00:08 +0100, floris wrote:
> codegentest failed because the foregroundlock was not(probably never)
> taken by the foreground thread than at some point it tried to
> temporarily release exactely this lock(in duchain.cpp), failing
> miserably. this patch makes the duchain.cpp function take the lock right
> before it tries to temporarily release if it hasn't already got it. it
> makes the test run, but is more of a hack than a fix...
> 
> Floris 

perhaps i should note that you need to foregroundlock in the testcleanup
(or what it's called?) of that test too to make it actually pass;