On 05/15/2012 07:56 PM, Paolo Carlini wrote:
But, speaking of incremental work: what if, post the build_min_nt_loc chunk, we handle build_min_non_dep and build_min in a case by case way? Thus we keep around the non-_loc variant and gradually replace each call? With testcases (small!) checking columns too.
That sounds fine.
Finally, I'm attaching my latest build_min_nt_loc patch. Turns out, in the light of the better understanding I now have, I can't justify most of the EXPR_LOC_OR_HERE, outside those for build_x_modify_expr which you explicitly indicated.
I think I was wrong when I indicated that, and that EXPR_LOCATION is better there, too. EXPR_LOC_OR_HERE is good for error messages, but not for setting the location of a tree. Though it occurs to me that we're likely to use the passed in location for errors as well, so we had better make sure we're passing in a useful location.
Jason
