Re: updating informix

Teruki Niki writes: > Did anyone experience the trouble with updating informix from V4.0 to V4.1. > I have some ESQL/C & 4GL applications of Version 4.0 (of satisfying quolity). > And I am anxious about the trouble caused by mismatch with V4.1. I'm preparing to make the same upgrade, and am testing all my code looking for problems. So far I've found one serious bug, and a couple of annoying glitches in RDS 4GL 4.10.UC1. (I can't address ESQL/C or c4gl, because I don't have them.) BUG 1) INPUT ARRAY goes into an infinite loop if the last field in the row is marked AUTONEXT, and if the user types enough characters to trigger the autonext. (If there's an AFTER ROW clause, it gets executed over and over, but it loops regardless of whether the The obvious workaround is to remove the autonext, which users can live without if they have to. I first ran into it in the middle of a complicated program, and it took an hour or so to isolate the problem :-( So I'm hoping to some everyone else some time ... This has been reported to Technical Support, and I've got a case number. Annoyances 2) The search path for source code doesn't work in fgldb 4.10.UC1. The USE command makes the change to the path correctly, but the debugger can't seem to find routines unless they're in the first directory in the path. I just sent mail to Tech Support about this, and don't have a case number yet. 3) You can't rely on "fglpc *.4gl". (This is much faster than letting your makefile do a separate fglpc on each source file, but ...) In previous versions doing "fglpc *.4gl" in some of my source directories used to crash (with a core dump), even though compiling the individual files would work correctly. That was annoying, but at least it was obvious when it failed. While testing my code with 4.10.UC1, I ran into a case where the compilation appeared to succeed, but the resulting failed subtly when executed :-( Recompiling the individual files fixed the problem. I haven't officially reported this: it doesn't seem worthwhile to send them a whole directory of code so they can try to reproduce it. But I thought I'd warn my fellow users :-) That's all so far, but I've still got more testing to do ... -- Harry Bochner -- bochner@das.harvard.edu