core dumps from 4GL programs
We have a situation where we are suddenly getting core dumps from one of our 4gl programs, which didn't used to have a problem. This only happens less than 1/2 of 1% of the time it is used. Sometimes it supposedly dies as soon as the user goes into it. Other times it dies in the middle. I have only been provided with one example with which I was able to re-create a core dump, and unfortunately, I blew the data away, when I ran it through the debugger (hoping to find an array being referenced out of bounds or some such folly which the debugger will catch for you.) But it worked w/ no problem. I have been able to determine that some of the core dumps seem to be coming from a DISPLAY <variable_name.*> TO <screen_record.*>. Any ideas on why this would happen? The system is an HP-UX 10.20 Informix 7.24.UC5 (Sorry, I don't know the version of the tools at the moment.) This is a production machine which normally has about 200-250 people logged in. I ran a file on all the cores I found and most of them were from this one program and they were mostly because they received a SIGSEGV. There were 1 or 2 other 4ge programs which had core dumps, but not enough so the users are complaining. I also noticed a core dump from logins, which seemed really strange to me. I couldn't find where we ran it, so I guess it's run as part of the login process. I'd appreciate any ideas. Thanks, Danny Wright danwright@bigfoot.com
Check C compiler programmers use. If GNU is used I beleive one should use the switch -fwritable-strings in order to work properly with screen records. Also check release notes for supporting compilers. HTH. Boris Niyazov Columbia Law School Systems Manager Danny Wright wrote: > We have a situation where we are suddenly getting core dumps from one of > our 4gl programs, which didn't used to have a problem. This only happens > less than 1/2 of 1% of the time it is used. > > Sometimes it supposedly dies as soon as the user goes into it. Other > times it dies in the middle. I have only been provided with one example > with which I was able to re-create a core dump, and unfortunately, I blew > the data away, when I ran it through the debugger (hoping to find an array > being referenced out of bounds or some such folly which the debugger will > catch for you.) But it worked w/ no problem. > > I have been able to determine that some of the core dumps seem to be > coming from a DISPLAY <variable_name.*> TO <screen_record.*>. > > Any ideas on why this would happen? > > The system is an HP-UX 10.20 Informix 7.24.UC5 (Sorry, I don't know the > version of the tools at the moment.) > > This is a production machine which normally has about 200-250 people > logged in. > > I ran a file on all the cores I found and most of them were from this > one program and they were mostly because they received a SIGSEGV. > > There were 1 or 2 other 4ge programs which had core dumps, but not > enough so the users are complaining. > > I also noticed a core dump from logins, which seemed really strange to > me. I couldn't find where we ran it, so I guess it's run as part of the > login process. > > I'd appreciate any ideas. > > Thanks, > Danny Wright > danwright@bigfoot.com