Re: Part II: Revenge of the CoreDumps

If the program is a compiled program, and is producing a core file, then you can start by getting the stack trace from the core file. This can be done by using xdb and/or adb. This should help in the resolution of the problem. What is the case number?? >Here we go again . . . > >4gl: 7.20.ud2 >OS: HPUX 10.20 >IDS: 7.23.uc11 > > >I've tested this particular program. I've beaten it up. I've added >displays. I've done everything but rewritten it. Here's the scoop. > >I tested many recommendations; here's a listing: > >Split functions Paul Watson tried -- no issue >String overflow Thomas Girsch possible -- I'll check again >Use debugger Thomas Girsch AAAAAAAAUGHH! Command line run works! >Initialize all variables before use Peter Ross checked - no issue. >Array checking with c4gl -a -anyerr David Williams checked - no issue. >Memory corruption ie string/array overflow June Tong testing numeric >overflow?? >Close window without first exiting from it June Tong NA as no windows in >batch program. >Input / display to forms David Madison displays only sqlca.sqlcode!= 0 >w/o using forms >Char / varchar > 255 characters? David Madison yes, but I don't get why >that would be a problem >Using GNU c-compiler?? Andrei S. nope -- standard C > >My particular program is a batch program using sysmaster. Other than >cleaning up the obvious ( CLOSing and FREEing cursors), I made one other >change. I noticed that a variable (p_partnum) was defined as an >integer. After all, it's defined that way in sysmaster.sql. So, I >changed it to a float. No problems on the command line after that. I >copied the 4ge to the appropriate directory and allowed cron to run the >given shell script as normal. It ran for 5 days with no problems, but >it blew up early this morning. > >I then ran it from the command line; no problems. >I reset the time for the cron job; no problems. > >Any thing else? I have a call open to Informix, but no luck so far. Madison Pruet