Corrupted Repository Indexes (OpenInsight 32-bit Specific)
At 17 DEC 2002 05:08:13PM Richard Guise wrote:
In case you too suffer this experience …
Found items disappearing from the file open lists in sys editor, screen painter, RDK and repository outliner.
Rebuilt system indexes (ran as normal), ran scan_rep (no problems), tore hair out and consulted gurus, oracles, soothsayers, high priests, etc.
Reconstituting repository record (even by copying onto itself) restored record to lists. On system index rebuild they vanished again. Same records each time.
Finally turned out to be several corrupted repository records with correct keys but badly corrupted contents. The visible clues were in sys editor Open Record listing of SYSREPOSIX where some records starting ATX and ATX had very obviously corrupted record keys. Opening such records crashed the system - so I had to hunt in the repository.
Once spurious records deleted, system index rebuild recovered the missing records in the lists.
Seems system index rebuild doesn't check or report on validity of what it's processing.
At 18 DEC 2002 06:26AM Oystein Reigem wrote:
Richard,
I got ATX and other rows with corrupted record keys while experimenting with international character sets:
ATX*0x7fffe 0xf0016408**1*$PUBLIC
AXX*0x7fffe 0xf0016408*1*$PUBLIC
TBA*0x7fffe 0xf0016408
The problem was caused by having @SVM= in the collation sequence. Revelation's got @SVM= included in their Create_Ansi_Sequence function for making international collation sequences.
Just in case this is what you've been doing to. Not very likely, perhaps, but the consequences are such you'd want to get rid of the problem rather sooner than later.
- Oystein -
At 19 DEC 2002 05:55AM Richard Guise wrote:
Thanks - but …
(1) I have never played with intl char sets in OI - so that cannot be the cause in my case.
(2) The corrupted keys I got looked a lot more poisonous than those you show. They seemed normal at first but with lots of gobbledegook appended. Two of the three crashed the system if one tried to look at them in sys editor.
I haven't been doing any crazy experimentation or tweaking in the repository or anything else strange - so I really have no idea what might have caused this. If I get any clues or ideas I'll post them.
It just worries me a little that the sys repos index rebuild clearly has no checks or warnings if it tries to process corrupted repos records. Should RTI add this in a future release as the problems only seem to become apparent when items get missed off lists, rebuilding can make things worse rather than better and finding what's happened isn't obvious or for the faint-hearted?