Re: sticky rollback...

> [bla bla bla] > > > >The problem is that under varying circumstances, the program will just > >hangup on a 'rollback work'. When we look at the process, the fglgo is > >sucking time, but the engine is quiet. > > > > The symptoms sound similar to a known bug, which shows up in some rare > usages: > > Bug 11776: > Short Desc [BEGIN-INSERT-ROLLBACK HANGS IN RDS, EXECUTES SUCCESSFULLY IN I4GL] > Long Bug Description > [A fairly simple transaction (BEGIN-INSERT-ROLLBACK) hangs during ] > [the ROLLBACK WORK statement under RDS, but works as expected ] > [when the code is compiled under I4GL. ] > [ ] > Bug Comments > [Tested 4.00.UE1 and .UJ1, 4.10.UC1: all work in 4GL but fail RDS ] Yup! That's got to be it. After all the futzing, we pretty much decided that it was the 4GL, and not the database. We did, however, end up "fixing". At least we found a work around. Basically, what we did was take most (maybe all) of the prepared statements we were using, and reverted them back to just generic SQL in the 4GL. One particular routine was doing all of the damage, but that routine works EVERY where else, so I don't know why it picked our particular program to barf on. We kept those statements that were required (Like an OPEN <cursor> USING x,y,z), but all of the others were just replaced. When all was said and done, that seemed to fix the problem. Later in the week, I'll bundled up all of the tables, the program, and the sample data required and ship it on up so you folks can pick it apart. Will (uunet!la4gen!villy)