GNUCobol compiled program counts one more record than expected

Question

I'm learning COBOL programming and using GNUCobol (on Linux) to compile and test some simple programs. In one of those programs I have found an unexpected behavior that I don't understand: when reading a sequential file of records, I'm always getting one extra record and, when writing these records to a report, the last record is duplicated.

I have made a very simple program to reproduce this behavior. In this case, I have a text file with a single line of text: "0123456789". The program should count the characters in the file (or 1 chararacter long records) and I expect it to display "10" as a result, but instead I get "11".

Also, when displaying the records, as they are read, I get the following output:

(There are two blank spaces between 9 and 11).

This is the relevant part of this program:

FD SIMPLE.
01 SIMPLE-RECORD.
    05 SMP-NUMBER PIC 9(1).

[...]
PROCEDURE DIVISION.

000-COUNT-RECORDS.
    OPEN INPUT SIMPLE.
    PERFORM UNTIL SIMPLE-EOF
        READ SIMPLE
            AT END
                SET SIMPLE-EOF TO TRUE
            NOT AT END
                DISPLAY SMP-NUMBER
                ADD 1 TO RECORD-COUNT
        END-READ
    END-PERFORM

    DISPLAY RECORD-COUNT.
    CLOSE SIMPLE.
    STOP RUN.

I'm using the default options for the compiler, and I have tried using 'WITH TEST {BEFORE|AFTER}' but the result is the same. What can be the cause of this behavior or how can I get the expected result?

Edit: I tried using an "empty" file as data source, expecting a 0 record count, using two different methods to empty the file:

$ echo "" > SIMPLE

This way the record count is 1 (ls -l gives a size of 1 byte for the file).

$ rm SIMPLE
$ touch SIMPLE

This way the record count is 0 (ls -l gives a size of 0 bytes for the file). So I guess that somehow the compiled program is detecting an extra character, but I don't know how to avoid this.

possibly (likely) you're reading the end-of-line character as 11th char. If you print each char as you read it, you could tell. — geert3, Dec 17 '17 at 12:52
Note: When using GnuCOBOL I'd recommend to use `PERFORM FOREVER` or `PERFORM UNTIL EXIT` and do an `EXIT PERFORM` in the `AT END` clause. — Simon Sobisch, Dec 17 '17 at 13:23
@geert3, I have added a DISPLAY statement, and the output is a bit baffling. — jantoniomartin, Dec 17 '17 at 15:17
I did as @SimonSobisch suggested and the result is just the same. Also, there isn't a newline character in the file. — jantoniomartin, Dec 17 '17 at 15:22

jantoniomartin · Answer 1 · 2017-12-19T13:53:05.633

1

I found out that the cause of this behavior is the automatic newline character that vim seems to append when saving the data file.

After disabling this in vim this way

:set binary
:set noeol

the program works as expected.

Edit: A more elegant way to prevent this problem, when working with data files created from a text editor, is using ORGANIZATION IS LINE SEQUENTIAL in the SELECT clause.

Since the problem was caused by the data format, should I delete this question?

edited Dec 19 '17 at 13:53

answered Dec 17 '17 at 19:28

jantoniomartin

215
3
9

1

You should not delete this question, because it may help future visitors that encounter the same problem. – Dec 25 '17 at 11:45

GNUCobol compiled program counts one more record than expected

1 Answers1