Re: [PATCH] Fortran -- Handle BOZ in accordance with F2008/2015.

[Steve Kargl <sgk at troutmask dot apl dot washington dot edu>]

On Fri, Oct 06, 2017 at 09:30:06PM -0700, Jerry DeLisle wrote:
> On 10/06/2017 01:34 PM, Steve Kargl wrote:
> > All,
> > 
> > I have spent the last few days trying to reconcile the various Fortran
> > standards' requirements for handling BOZ.  The short story is that J3
> > over the last 27 years has made incompatible changes to the interpretation
> > of a BOZ (under some circumstances).  The interpretations in F2008 and
> > F2015 now treat a boz-literal-constant as a sequence of bits.  Unfortunately,
> > due to quirks in how BOZ are currently implemented in gfortran and a boat
> > load of documented and undocumented extensions, bringing gfortran into 
> > agreement with F2008/F2015 led to a rewrite of BOZ handling.  In the 
> > rewrite I have made no attempt to use the -std= option to try to maintain
> > the incompatibilities between standards.
> > 
> 
> The patch tests OK on trunk with linux-x86_64. I suggest this be
> committed as is and then see if there is any fallout that can be
> dealt with -std=legacy or similar later.
> 

Thanks, but unfortunately in my review of the patch and F2008/F2015
found an issue. 

  integer :: i = int(z'123456789',4) 

isn't handled correctly.  We do get a silly error.

% gfcx -c a.f90
a.f90:1:19:

 integer :: i = int(z'123456789',4)
                   1
Error: Arithmetic overflow converting INTEGER(-1) to INTEGER(4) at (1).
This check can be disabled with the option '-fno-range-check'

F2008 13.3.3 has

    When a boz-literal-constant is the argument A of the intrinsic
    function INT or REAL,
    ...
    if the length of the sequence of bits specified by A is greater
    than the size in bits of a scalar variable of the same type and
    kind type parameter as the result, the boz-literal-constant is
    treated as if it were truncated from the left to a length equal
    to the size in bits of the result.

    C1301 If a boz-literal-constant is truncated as an argument to
    the intrinsic function REAL, the discarded bits shall all be zero.

We handle C1301 correctly except the error message could be better.

% gfcx -c a.f90
a.f90:1:17:

 real :: i = real(z'123456789',4)
                 1
Error: BOZ constant at (1) is too large (33 vs 32 bits)

-- 
Steve
20170425 https://www.youtube.com/watch?v=VWUpyCsUKR4
20161221 https://www.youtube.com/watch?v=IbCHE-hONow