Hi Jerry, I am not sure that I understand why my suggested patch "threw Claude into a tizzy", The two chunks that you pushed are identical. My version turned one into a function in trans-array.cc, provided a prototype and called it from trans-expr.cc, thereby halving the size of the patch.
I have learned now that you really have to take a close look at Claude-generated patches. They tend to code and comment bloat. Also, take care with respect to pointers to the standard. It seems to generate them at random. Refactoring your patch is on my todo list and I will apply it after you backport. Regards Paul On Thu, 4 Jun 2026 at 20:52, Jerry D <[email protected]> wrote: > > On 6/3/26 11:54 PM, Paul Richard Thomas wrote: > > Hello Jerry, > > > > Contrary to my remark on Mattermost, this looks good to me except that > > I would halve the size of the patch by refactoring it as in the > > attached version. It passes regression testing like the original. > > > > OK for mainline and backporting later on. > > > > Thanks for re-examining this Golden Oldie! > > > > Paul > > Paul, the diff you attached threw claude into a tizzy. I slapped it and set it > back on track. The attached fixes the breakage in the original > assumed_rank_7.f90 with -fsanitize=address, passes the new > asan/assumed_rank_26.f90 and fixed a breakage in coarray/dummy_3.f90. > > I have completely regression tested this one and manually tested the original > assume_rank_7.f90 with -fsanitize=address. All pass. > > I plan to commit this shortly. > > Regards, > > Jerry
