Hi Friends
I recently installed an upgrade for a client from R:Base for DOS to R:Base
7.1 (yes, I know - this was long delayed and they were unwilling to go for
7.6 yet. That will be next.).
As part of the upgrade, I added several PK / FK relationships where in the
old version they had used just indexes. When they ran my automated check
the database / pack keys / backup routine, the new constraints "broke". The
data itself was fine, but if you did a:
SELECT * from table2 WHERE FKcolumn NOT IN (SELECT PKColumn from Table1),
ALL the rows in table 2 came back.
I believe I've fixed this by doing an UNLOAD ALL to file, and INPUT to
recreate the database.
But the real issue is how they discovered the problem.
When the program hit the PACK KEYS command, R:Base found the broken index
AND STOPPED the pack keys process. They got the constraint error message
about invalid FK on the screen, even though ERROR MESSAGES and MESSAGES were
both set OFF. But what they did NOT get was my PROGRAMMED message that the
PACK KEYS had failed and to contact me. The program instead merrily went on
its way assuming the database was OK and backed it up.
Here's the code that was failing:
SET ERR VAR gError
PACK KEYS
SET VAR vErrHold = .gError
IF vErrHold <> 0 THEN
-- code for error message
GOTO BOTTOM
ENDIF
The error message the DID get - the constraint violation message - of course
made little sense the user in the context of the backup process, but
fortunately he was sharp enough to know something was up and to call me.
When the indexes were broken, and a PACK KEYS failed, the error variable at
the R> prompt was NOT 0, but when I traced the program, it was right after
the PACK KEYS
So 2 questions:
1. Is this is a known problem in 7.1 and has it been addressed in later
versions?
2. If this is NOT a bug, and R:Base is behaving correctly, is there
another way to trap for the issue in a program and give the user a
customized error message?
David Blocker
David M. Blocker
[EMAIL PROTECTED]
781-344-1920
Cell Phone: 339-206-0261
