Bug in PrioXRef

I haven't had the need to touch the source files of this program¹ since 12/21, although I have to admit that I haven't been using it very much lately. Actually, that's not true: I used it to find bugs in a very convoluted procedure that someone else had written years ago: it turns out that the value of RETVAL was never checked after opening about ten cursors. The bugs disappeared once the value of RETVAL was checked.

The other day I wanted to run the program on a procedure that was causing me problems; was the variable that I needed called :$.M0, :$.N0 or even :$.P0? PrioXRef got stuck in an infinite loop when checking the procedure, so even this tool couldn't help. Eventually I realised what the problem was: those three variables come from three different CHOOSE stages, but only one of them will actually be chosen. In the end, I added a line in three different stages to save the value of whichever variable was current into a fourth global variable. But that's not the point.

The procedure was about 190 lines long, but the error seemed to occur in one of the final lines. Eventually I saw what construct was causing the infinite loop: a line that had two comments on it. Why two? Because I had commented out the first part, but the original comment was still extant. 

/* EXECUTE SENDMAIL :$.MBX, :XMSG; */ /* send the letter! */

Finding the offending construct was half of the problem, but I also had to discover why PrioXRef was entering an infinite loop. First, I created a simple three line text file, where the second line had the double comment. Then I single stepped through the program code a few times until the penny dropped and I saw where the bug in my code was.

  if (token = '/') and (ch = '*') then
   begin   // comment
    repeat
     n:= pos ('*/', buffer); 
     if n = 0 then
      begin
       bufptr:= buflen + 1;
       getchar;
      end
    until n > 0;

The problem is with the call to the function pos; this always starts at the beginning of the string (named buffer) and so the procedure kept on finding the closure of the first comment. The simplest way to handle this was to delete everything from the beginning of buffer until the closing backslash; I also had to reset the variable that contains the length of buffer (so that the program will know that it has reached the end of a line and has to start operating on a new line). The reason for the repeat/until loop is that the comment may continue on a second line and so the terminator won't be found; setting bufptr (the index into the buffer) to the length of the buffer plus one will ensure that the getchar procedure will read a new line.

Once I had this fixed, I could now run PrioXRef on the procedure that I wanted to check.

Internal links
[1] 1338

---

This day in history:

Blog #	Date	Title	Tags
452	13/02/2012	Gateway	Gateway
673	13/02/2014	A flaw with spreadsheets	ERP, DBA, Excel
926	13/02/2016	ERP thoughts	DBA
1583	13/02/2023	Putting words into action	Israel