Sanity check

[William J. McEachran]

On Tue, 2 Jun 2020 at 12:10, Fairlight via Filepro-list <
filepro-list at lists.celestial.com> wrote:

> It's the result of bad BTree+ code.
>
> If you go through this demo, you'll see when data needs to be split.
> Something about it is incorrectly handled in fP, resulting in bad
> indexes.
>
> https://www.cs.usfca.edu/~galles/visualization/BPlusTree.html
>
>
I got lucky and found that the nightly tar archive  had the troubled data.
I was able to play around with it on a test system and found I could get a
DKNF (deleted key not found) error on the data in addition to the sanity
check
The data is part of a simple routine that is created and used in a single
process ... so it easily traceable.

All lookups involved in record operations (creating, deleting, modifying)
are protected (-p flag) lookups.
Each record operations concludes with a write - 'write [lookupname''.
 Create a record; write that record.  Delete a record; write that record.
etc.
The routine concludes with a 'sync'.

So I can't think of anything that I've missed to cause invalid key/index
issues.
Did I miss something?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.celestial.com/pipermail/filepro-list/attachments/20200602/7d8f4995/attachment.html>