------- Comment #17 from damian at rouson dot net 2010-07-31 02:30 ------- Tobias,
Thanks for your continued efforts on this. It will ultimately have a huge impact on the usability of gfortran for my purposes. I look forward to hearing more when you get back to it or when others do. Since there are regressions, I assume I won't get access to your patch just by doing an update using macports (my preferred method for building), but I look forward to trying it out once it's available. Damian (In reply to comment #16) > Remove assignment. I think I won't work on this in the next weeks and "New" is > better in allowing others to pick it up. If not, I will look at it again > later. > > The patch of attachment 20714 [edit] plus the fix in comment 15 (plus the > test cases > at the bottom of attachment 20696 [edit]) mostly work OK - except for a > handful > regressions. The implementation was using "DT-name" as generic function and > created the derived type gfc_symbol using the suffix "@". > > It was suggested to do the following - which might be cleaner: > - Create the <dt name> as generic function (as before) > - Use ns->derived_types to save the derived types (instead of generating > another symbol) > > I think one can mostly recycle the resolve.c part of the patch - and the > gfc_convert_to_structure_constructor's part in primary.c (one can probably get > rid of gfc_match_structure_constructor after some modifications). However, all > the rest has to be adapted. > -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39427
