#874 DISPLAY for variables with `P` picture may output one byte too much / leading zeroes

[Denis HUGONNARD-ROCHE]

Hi
May by a bug in display statement with P Picture
See the program attached

[Simon Sobisch]

Yes, that's a bug on the DISPLAY side:

• cob_display iterates over all real and internal fields via varargs
• it gets both real field with their separate data-pointer ("01" for the first after MOVE, unchanged "00"for the second field) and attribute
• those also get placed to pretty_display_numeric()
• this does correctly setup a temporary field for display
• the temporary field has a fixed data buffer, which is not zeroed after use (so may contain the last data)
• the "internal numeric edited" second field is moved to the temporary one vie cob_move_display_to_edited(), which fills the data up to its digits (4)

the issues here are:

• cob_move_display_to_edited() only stores data up to the PICTURE (all pic should match the size of edited fields but in this case this isn't true)
• pretty_display_numeric() only takes positives scales into account for the size calculation

I'm working on a fix (along with refactoring to and use similar code as used in other places).

[Simon Sobisch]

Minimal reproducer of the original issue:

       IDENTIFICATION DIVISION.
       PROGRAM-ID. prog.
       DATA DIVISION.
       WORKING-STORAGE  SECTION.
       01 P1 PIC     999 value 1.
       01 P2 PIC     9P  value 0.
       PROCEDURE DIVISION .
           display '00=' P2 '; 001=' P1 '; 00=' P2 '.'
           GOBACK .

With every GC version tested (2.2+ up to 3.2rc2) that results in:
00=00; 001=001; 00=001. with -fpretty-display, which is the default and in
00=00; 001=001; 00=00. with -fno-pretty-display.
If there is any P after the data then it prints one byte too much "whatever is in the internal terminal buffer".

Extending the test showed that on the other hand -fno-pretty-display has another bug, it prints the amount of Ps as leading zeroes (so 100 in 9PP gets displayed as 001).

Both fixed with [r4998].

Fair warning: there are a bunch of pending issues with P data fields, I even consider to explicit mark this PICTURE character as unfinished...