[Patch?, fortran] PRs 31197, 31258 and 31879 - character things

[Paul Richard Thomas]

Dear All,

Being aware that the stage 1 cut-off is with us, I wanted to know what
you would like me to do with the attached, noting that:

(i) It bootstraps and regtests on x86_ia64/FC5;
(ii) It fixes the above PRs and more besides;
(iii) It represents a small part of my original ambition - whilst I
have tested the removal of large chunks of _conv_ and _trans_ code,
associated with character fixups, this still causes some regressions;
(iv) I am not convinced that the frontend and resolve handling of
characters is where it shoul be, yet; and
(iv) Complex substrings with variable expressions in substring
references are still not 100% fault free; eg.

in the last line of the attached char_length_6.f90, changing
if (any(ccopy(z//mz(:)(2:3)) .ne. (/"zzgh ","zzjk "/))) call abort ()
to
  i = 2
if (any(ccopy(z//mz(:)(i:3)) .ne. (/"zzgh ","zzjk "/))) call abort ()

causes a segfault in the same place as PR31217

In summary, it fixes stuff, probably does no harm but is a long way
from complete.

Cheers

Paul