https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123693

--- Comment #8 from anlauf at gcc dot gnu.org ---
(In reply to Steve Kargl from comment #7)
> Created attachment 63431 [details]
> Patch file
> 
> The attached patch does two things.  First, it updates
> gfc_check_ltime_gmtime()
> in check.cc to check that the arguments of ltime() and  gmtime() have the
> same
> type. The type is default integer kind, which agrees with the documentation
> for gfortran 3.4.6.  Second, I have updated intrinsic.texi so that the manual
> makes it clear that the expected type is default integer kind and both
> arguments
> must have this type.  Thus,
> 
>   integer tm
>   integer val(9)
>   ...
>   call gmtime(tm, val)
> 
> will work with and without the -fdefault-integer-8 and -finteger-4-integer-8
> option.
> 
>   integer(4) tm      ! or vice versa
>   integer(8) val(9)
>   ...
>   call gmtime(tm, val)
> 
> will raise an error, i.e.,
> 
> % gfcx -o z foo.f90 
> foo.f90:5:20:
> 
>     5 |     call gmtime(tm, val)
>       |                    1
> Error: 'values' argument of 'gmtime' intrinsic at (1) must be the same
> type and kind as 'time'

We should also reject any kind /= 4 and /= 8 as unsupported.

Reply via email to