This month, I would like to review the ICIs that we have had for a few releases plus those that have recently appeared, and then the trouble with twelve …
It began …
Db2 development realised that something had gone horribly wrong when a bunch of Db2 users suddenly found that the output from their queries was no longer what it should be … After a bit of digging, the CHAR format rewrite was found to be the root cause and the fix was hastily created – BIF_COMPATIBILITY ZPARM.
What’s in a name?
Well, then VARCHAR happened and along came a very unpleasent problem with JAVA timestamps and, as I have documented in earlier BLOGs, it all started getting silly with one ZPARM being used for multiple format problems.
Along came the ICI (Incompatible Change Indicator)
So in Db2 10 we got a new IFCID, the 366, which was spat out at *every* prepare (bind) of any SQL that, possibly, contained an ICI. Now we started off pretty small with just three ICI’s: the first two being the reformatted output of CHAR and VARCHAR and the third being the TIMESTAMP format problem.
Db2 10 updates for Db2 11
Here they brought out numbers four to nine to handle all the little changes in Db2 11 so that you got the alert in Db2 10 before it bit you in Db2 11 – all well and good.
Db2 11 updates
The big change, was the brand new IFCID 376 – which is the evil twin of the 366. The only difference being that Db2 cached the entries, so you basically got a rolled up 366 – apart from one tiny little detail. The Execution count was missing. For the 366 it is 1:1, but for the 376 it is 1:nnnnn which could be any positive integer. They then added the 11nn range, going all the way up to 1111, and then they brought in 1112 for empty XML tags. Now all of these have been discussed in my earlier blogs.
What’s new in the ICI World?
Db2 12 of course! They brought out 1201 very early on due to POWER causing a problem. The output on overflow changed from a negative to a positive SQLCODE, which can of course cause “problems”… Why did this change even happen? IBM rewrote the code from using LE 32 bit assembler math calls to using C, and so the function “knew” if it overflowed and could return a warning saying so, whereas the 31 bit assembler just died a death and you got a negative SQLCODE.
Then it went quiet for a while until something weird happened: 1215031 and 1215032 appeared. Now, at first, I liked the idea of putting the FL into the ICI, but then I realised it was actually pretty pointless and just made it more confusing !
1215031 is issued when you could qualify a row with NULL in the DATA CHANGE OPERATION column using the FOR SYSTEM_TIME FROM/BETWEEN predicate on a system period temporal table with AUDITING.
1215032 is issued when you attempt to call stored procedure SYSPROC.SET_MAINT_MODE_RECORD_NO_TEMPORALHISTORY as this is no longer supported for data replication calls.
1204 (note the FL has gone…) is issued when you use CURRENT_SERVER or CURRENT_TIMEZONE as a column or variable name.
In the Docu it says 1202, but you actually get a 1215031 – and there’s no mention of 1215032 and 1204!
More new ones
Meanwhile, not (nothing?) to do with Db2 12, IBM also brought out
11 for using SELECT INTO syntax with a UNION
[12 was thankfully skipped!]
13 for INSERT/UPDATE/DELETE using an attribute WITH UR
Both of these were more parsing bugs than anything evil, but both require code changes if they appear!
Now IBM have enhanced the LOAD utility with LOAD FORMAT DELIMITED for correct packed numeric data support when one or more virtual decimal digits exist. This has caused another problem, as the data loaded could be viewed (as it was by one of our customers), as inconsistent. IBM then created another APAR to roll back the change and introduce a new ZPARM LOAD_DEL_IMPLICIT_SCALE to control how these numbers should be loaded. Default is NO, like it used to be, with an implied decimal point at the far right of the data. At the same time, it will now alert users that they could have an incompatibility with the LOAD by changing IFCID 25 to set a new bit. This warns you that you have done a LOAD into a table where there is packed decimal data with one or more digits after the virtual decimal point. If YES, then load interprets the Scale setting in the LOAD statement. For details please see APARs PH28104 and PH36908.
Pain Point for you?
The above mentioned new notification is a bit strange (pardon the pun), as there is already the IFCID 376 for incompatibilities. Now you must also start the IFCID 25 and go checking bits. So my question to you all is: Have you got this problem and, if so, do you think it is worth it to integrate IFCID 25 bit checking into SOFTWARE ENGINEERINGs/SEGUSs current ICI/BIF Use Case in our WorkLoad Expert and/or our BIF/ICI Freeware software?
The future is bright
As far as ICIs are concerned they just keep on rolling!
As always if you have any comments, especially with regard to IFCID 25, please feel free to e-mail!