https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126821
Robert Dubner <rdubner at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |rdubner at gcc dot gnu.org
--- Comment #1 from Robert Dubner <rdubner at gcc dot gnu.org> ---
I see that you found that I (I'll use "me" as a shortcut for "gcobol") am
storing, moving, and using P-scaled variables properly.
So, it became a question of what to do for DISPLAY.
I am not aware of solid documentation about DISPLAY.
So, in general, I operate from the principal of "make it human readable".
Consider a PIC S99V99. In an ASCII implementation, MOVE -12.34 to PIC S99V99
results in four bytes in memory: "123t". But in a gcobol compilation,
DISPLAYING that value produces the output "-12.34".
So, I do the same kind of thing for a PIC 999PPP. After a MOVE 123000 to that
value, the image in memory is "123". The result of a DISPLAY is "123000".
It is true for both variables, when they are part of a group, that displaying
the group shows the actual underlying character storage.
But I chose to display the S99V99 as "-12.34", even though the minus sign and
the decimal point are not part of the storage. And I chose to display "999PPP"
as "123000", even though the zeroes are not part of the storage.
Is this a problem? If it doesn't match historical IBM behavior, we could make
it part of the "-fdialect ibm" option.
