GnuCOBOL / Bugs / #1008 Regression in move to numeric edited items

Chuck Haatvedt - 2024-11-22

David,

This appears to be a bug in both the earlier revision as well as the current version.

from IBM doc...

https://www.ibm.com/docs/en/cobol-zos/6.3?topic=clause-floating-insertion-editing

from 2022 ISO standard pg 449

B Each symbol 'B' represents a character position that will be checked to contain — the character space if the symbol 'B' is neither part of floating insertion editing nor of zero suppression with replacement editing; — otherwise, the character space or, if no significant numeric character appears to its left, the corresponding floating insertion character or replacement character respectively.

here is the result with the Fujitsu COBOL compiler

F:\AA-minGW32-static>prog-1008 4 -4

with my fix...

F:\AA-minGW32-static>cobcd -x prog-1008.cbl F:\AA-minGW32-static>prog-1008 4 -4

ran my build from source with this patch and it passed all the testsuite and NIST tests successfully

note that this test program should be added to the testsuite to ensure that going forward we check for regression on this.

Chuck Haatvedt

Last edit: Chuck Haatvedt 2024-11-22

move-patch.txt
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

David Declerck - 2024-11-22

Hi Chuck,

Thank you for this patch ; that was quick ;)
I'll check that it passes the CI on our GitHub mirror (P.S: it's open to anyone, feel free to use it to check your patches).
Just to be more future-proof: can you think of any other "corner cases" of these moves to numeric edited that would be missing a test case ?
Also, checking with @sf-mensch: is this patch OK for you ?

David

Last edit: David Declerck 2024-11-22
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
- Simon Sobisch - 2024-11-22
  
  So NIST is working fine with it?
  
  Am 22. November 2024 14:28:08 MEZ schrieb David Declerck ddeclerck@users.sourceforge.net:
  
  Hi Chuck,
  
  Thank you for this patch ; that was quick ;)
  I'll check that it passes the CI on our GitHub mirror.
  Also, checking with @sf-mensch: is this patch OK for you ?
  
  David
  
  [bugs:#1008] Regression in move to numeric edited items
  
  Status: open
  Group: GC 3.x
  Created: Thu Nov 21, 2024 04:14 PM UTC by David Declerck
  Last Updated: Fri Nov 22, 2024 07:44 AM UTC
  Owner: nobody
  
  Hi,
  
  It appears revision 5290 caused some regression in moves to numeric edited items.
  
  Consider the following program:
```
IDENTIFICATION DIVISION.
PROGRAM-ID. prog.

DATA DIVISION. WORKING-STORAGE SECTION. 01 W-X1 PIC S9 VALUE 4. 01 W-X2 PIC -B- VALUE SPACES. 01 W-Y1 PIC S9 VALUE -4. 01 W-Y2 PIC +B+ VALUE SPACES. PROCEDURE DIVISION. MOVE W-X1 TO W-X2. DISPLAY W-X2. MOVE W-Y1 TO W-Y2. DISPLAY W-Y2. STOP RUN.

With 5290, it outputs:
-4
+4
While with any earlier revision, it outputs:
4
- 4
```
  
  I tried to have a look at the new optimized_move_display_to_edited function, but I believe the original author (@chaat) has a much better understanding of this code.
  
  Sent from sourceforge.net because you indicated interest in https://sourceforge.net/p/gnucobol/bugs/1008/
  
  To unsubscribe from further messages, please visit https://sourceforge.net/auth/subscriptions/
  
  Related
  
  Bugs: ~~#1008~~
  
  alternate
  
  If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
- Chuck Haatvedt - 2024-11-22
  
  here is from the IBM documentation, I've not checked the 2022 draft standard...
  
  Any simple-insertion symbols (B 0 / ,) within or to the immediate right of the string of floating insertion symbols are considered part of the floating character-string. If the period (.) special-insertion symbol is included within the floating string, it is considered to be part of the character-string.
  
  the comma works, but I've not checked the other two, the zero and slash.
  
  Chuck Haatvedt
  
  If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
- Chuck Haatvedt - 2024-11-22
  
  IDENTIFICATION DIVISION. PROGRAM-ID. prog. DATA DIVISION. WORKING-STORAGE SECTION. 01 W-X1 PIC S9 VALUE 4. 01 W-X2 PIC -B- VALUE SPACES. 01 W-Z2 PIC -//00BB,,- VALUE SPACES. 01 W-Y1 PIC S9 VALUE -4. 01 W-Y2 PIC +B+ VALUE SPACES. PROCEDURE DIVISION. MOVE W-X1 TO W-X2. DISPLAY W-X2. MOVE W-Y1 TO W-Y2. DISPLAY W-Y2. MOVE W-X1 TO W-Z2. DISPLAY W-Z2. STOP RUN.
  
  gnucobol with current patch...
  
  F:\AA-minGW32-static>prog-1008 4 -4 //00 4
  
  with Fujitsu compiler
  
  F:\AA-minGW32-static>prog-1008 4 -4 4
  
  So perhaps this is another existing bug....
  
  Chuck Haatvedt
  
  If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
  - Chuck Haatvedt - 2024-11-22
    
    here is the result with another patch...
    
    F:\AA-minGW32-static>COBCD -x PROG-1008.CBL F:\AA-minGW32-static>prog-1008 4 -4 4
    
    This now matches the output from the Fujitsu compiler.
    
    i've attached a new version of the patch file.
    
    Last edit: Chuck Haatvedt 2024-11-23
    
    move-patch.txt
    
    If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

also testsuite #532 needs to be fixed as it is incorrect

#                             -*- compilation -*-
532. run_fundamental.at:450: testing MOVE to edited item (4) ...
./run_fundamental.at:2122: $COMPILE -Wno-truncate prog.cob
./run_fundamental.at:2123: $COBCRUN_DIRECT ./prog
--- /dev/null   2024-11-22 17:49:35.000000000 -0600
+++ /home/spcwh2/x32/gnucobol-trunk/tests/testsuite.dir/at-groups/532/stdout    2024-11-22 17:49:35.102903300 -0600
@@ -0,0 +1,2 @@
+EX-13.02 EXPECTING ==>'    $//00//10  '  WHAT I GOT WAS ==>'      $00//10  '<==
+------------------001 TESTS FAILED------------------
532. run_fundamental.at:450: 532. MOVE to edited item (4) (run_fundamental.at:450): FAILED (run_fundamental.at:2123)

the expected result should be

WHAT I GOT WAS ==>'      $00//10  '

here is the patch file for run_fundamental.at

Last edit: Chuck Haatvedt 2024-11-23

run_fundamental-patch.txt

David Declerck - 2024-12-06

Hi Chuck,

Seems there's a bit more that has been broken with these patches.
With 5285, BLANK WHEN ZERO no longer works in some cases.

IDENTIFICATION DIVISION. PROGRAM-ID. prog. DATA DIVISION. WORKING-STORAGE SECTION. 01 W-X PIC -ZZZBZZ9,99 BLANK WHEN ZERO. PROCEDURE DIVISION. MOVE 0 TO W-X. DISPLAY W-X. STOP RUN.

This displays:

0,00

The check for zero in optimized_move_display_to_edited might be a bit too restrictive, since the source data may contain a sign (in the example above, the literal 0 ends up in a display field which contains 00000000+).
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

David,

here is a patch file for MOVE.C

the issue was when checking for "BLANK WHEN ZERO" was that if the sending field is signed, then we need to scan for 1 less byte. Note that the intermediate field is always created as display numeric with the same number of digits and scale as the receiving picture field. Also if the receiving field is signed then the intermediate field will always be signed separate and sign trailing.

   IDENTIFICATION DIVISION.
   PROGRAM-ID. prog.
   DATA DIVISION.
   WORKING-STORAGE SECTION.
   01  W-X  PIC -ZZZBZZ9,99 BLANK WHEN ZERO.
   01  W-Y  PIC  ZZZBZZ9,99 BLANK WHEN ZERO.
   01  W-Z  PIC +ZZZBZZ9,99 BLANK WHEN ZERO.
   01  J    PIC S9(8)  COMP-5.
   PROCEDURE DIVISION.
       MOVE 0 TO J.
       MOVE J TO W-X.
       DISPLAY ' ===>' W-X '<==='.
       MOVE 1234 TO J.
       MOVE J TO W-X.
       DISPLAY ' ===>' W-X '<==='.
       MOVE -1234 TO J.
       MOVE J TO W-X.
       DISPLAY ' ===>' W-X '<==='.

       MOVE -1234 TO W-X.
       DISPLAY ' ===>' W-X '<==='.

       DISPLAY ' '.

       MOVE 0 TO J.
       MOVE J TO W-Y.
       DISPLAY ' ===>' W-Y '<==='.
       MOVE 1234 TO J.
       MOVE J TO W-Y.
       DISPLAY ' ===>' W-Y '<==='.
       MOVE -1234 TO J.
       MOVE J TO W-Y.
       DISPLAY ' ===>' W-Y '<==='.

       MOVE -1234 TO W-Y.
       DISPLAY ' ===>' W-Y '<==='.

       DISPLAY ' '.

       MOVE 0 TO J.
       MOVE J TO W-Z.
       DISPLAY ' ===>' W-Z '<==='.
       MOVE 1234 TO J.
       MOVE J TO W-Z.
       DISPLAY ' ===>' W-Z '<==='.
       MOVE -1234 TO J.
       MOVE J TO W-Z.
       DISPLAY ' ===>' W-Z '<==='.

       MOVE -1234 TO W-Z.
       DISPLAY ' ===>' W-Z '<==='.
       STOP RUN.


output...

F:\AA-minGW32-static>cobcd -x prog.cbl
F:\AA-minGW32-static>prog
 ===>           <===
 ===>      12,34<===
 ===>-     12,34<===
 ===>-     12,34<===

 ===>          <===
 ===>     12,34<===
 ===>     12,34<===
 ===>     12,34<===

 ===>           <===
 ===>+     12,34<===
 ===>-     12,34<===
 ===>-     12,34<===

This looks good to me, however you should confirm that and add this to the existing test in the testsuite.

  Thanks,

❤️

Last edit: Chuck Haatvedt 2024-12-06

move-bug-1008-patch.txt

Chuck Haatvedt - 2024-12-11

David,

are you working with Simon to implement the latest patch that I sent to resolve the "BLANK WHEN ZERO" issue with a signed field ?

Chuck Haatvedt
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Simon Sobisch - 2024-12-17

labels: --> BLANK WHEN ZERO, editing, libcob

status: open --> closed

assigned_to: David Declerck
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Simon Sobisch - 2024-12-17

Fixed with [r5403].

Related

Commit: [r5403]

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Regression in move to numeric edited items

A free COBOL compiler

Group

Searches

Help

#1008 Regression in move to numeric edited items

Related

Discussion

Related

Related