> Hello,
> 
> The code to push arguments onto the stack, for many ABI's, involves two passes:
> 
> 1. iterate over the arguments. compute the location/offset of each, accumulate results (to compute the size of the stack frame).
> 
> 2. iterate over the arguments, compute the location/offset of each, store accordingly
> 
> The attached patch rewrites arm_push_arguments() so that the two passes are combined into a single two-pass for-loop.  The only difference between the first and second pass of the loop is that the second pass stores values.  The intent is to make the maintenance of the function easier - e.g. adding FP support would now involve one addition (currently it involves two).
> 
> Thoughts, on the general approach?
> 
> It also adds the command ``set debug arm'' which will cause the code to print out the arguments as they are pushed.
> 
> Ok for arm?
> Andrew 

Hmm, with patch ...