[Bug fortran/102314] [12 Regression] ICE in verify_ssa, at tree-ssa.c:1076 since r12-1319-gd4d38135b3137f1d

["jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102314

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |burnus at gcc dot gnu.org

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I think those types:
  character(kind=1)[1:.y] * y;
don't really make sense for the character(:) allocatables, I think much better
would be to leave the TYPE_MAX_VALUE from TYPE_DOMAIN out, making it [1:].
[1:.y] nicely expresses the intent, in most places y (if it is not NULL) should
point to [1:.y] array.  But if we emit a DECL_EXPR for it, it will be just
wrong,
the middle-end expects the C/C++ VLA behavior, the size is computed once (when
encountering the DECL_EXPR), then the VLA is allocated with it and that is the
size it has from that point on.
The deferred-length (is that the right term) arrays work differently from this
though, the length can be changed at any time.  So the VLA length is always
determined from the current value of the .y variable (except in the short spots
where that is updated and the var not yet reallocated).