On Sun, Jan 21, 2018 at 8:30 PM, Steve Kargl <s...@troutmask.apl.washington.edu> wrote: > On Sun, Jan 21, 2018 at 01:11:28PM +0200, Janne Blomqvist wrote: >> PING >> > > Janne, > > I didn't see anything wrong with the patch. Other than Dominiq's > concerns about slow downs with -m32, I think it is fine.
Thanks, I guess that's an Ok? I'll rebase on the latest trunk, run my tests again and commit. I think those issues Dominique mentioned are due to wraparound with the old code, so now it's at least correct even if slow. > I do wonder why you use gfc_charlen_t in some place and > HOST_WIDE_INT in other places as gfortran.h defines > gfc_charlen_t to be HOST_WIDE_INT. gfc_extract_hwi takes a pointer to a HOST_WIDE_INT, so if we at some point change the typedef for gfc_charlen_t that would break. Also, similarly, gfc_mpz_get_hwi returns a HOST_WIDE_INT so I thought it's better to use that type than converting to some other which might or might not be the same. Similarly, for things like array indices in loops etc. I preferred size_t. -- Janne Blomqvist