[Bug fortran/84734] [8 Regression] Compiling codes with insane array dimensions gives an ICE after r257971

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84734

kargl at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|6.4.0, 7.3.0                |
           Severity|normal                      |minor

--- Comment #4 from kargl at gcc dot gnu.org ---
(In reply to Dominique d'Humieres from comment #3)
> Up to revision r257970 compiling the code in comment 2 gives there errors
> 
> size_kind_2_db.f90:6:13:
> 
>  integer :: B(huge(1_8)+3_8,2_8)
>              1
> Error: Arithmetic overflow at (1)
>

Tracing through gdb shows that error is queued.
It simply isn't emitted.  Have no idea why.
I tried using gfc_error_now() and the error
was still not emitted.  I'll see if I can find
the root cause, but this way down my list of
important bugs.

I've removed the "known to work" entries
as I get failures on FreeBSD.