This is targeted at Adam, but sharing because it's caring… There's a bit of awkwardness in our model right now in dealing with converting the logical “inputs” into the compilation to real instructions.
At the moment, this is spread between the CompileSpec and the Compiler implementation. The ideal situation would obviously be if this was solely the responsibility of the Compiler (i.e the spec is a declarative abstraction that the compiler impl interprets into its language). However, this is complicated by the fact that there are many different potential compiler settings that we won't be able to model via the spec. The only idea I have is for a hybrid model where we model what we can, but allow the user to specify arbitrary raw args as part of the spec as well. This is effectively what we are doing now, but just not very well. I'm not proposing changing anything here pre 1.0, this is just something I came across adding GCC3 support and raised for discussion. -- Luke Daley Principal Engineer, Gradleware http://gradleware.com --------------------------------------------------------------------- To unsubscribe from this list, please visit: http://xircles.codehaus.org/manage_email
