[Resolved?] Re: ovsqlite - bug?

[Julien ÉLIE]

Hi Kamil,

> Well, it seems that problem is with spool not with sqlite.

Oh, great, thanks for your investigation!

ovsqlite appears to be more robust than ovdb as it did an integrity 
check before accepting new overview data (maybe by primary key in 
database, I have not checked).


> It is worth noting, that greping /var/lib/news/history for these storage
> tokens returns nothing.
> 
> So the first thigs I should do is to make history consistent with spool
> and then test further.

I don't think you should bother.
You won't be able to make your history file consistent with your 
self-expiring CNFS buffers.  I bet the articles have already expired 
from your current overview.  They are just still here in the buffer, 
just waiting for being overwritten by the next cycle.

What happens is that makehistory parses your spool and finds these old 
articles.  It generates overview entries for them.
Tomorrow, after the run of expireover (by news.daily), your overview 
date will be purged of these entries.


> But for now I am pretty sure, that sqlite worked properly as, it forces
> uniqueness of  (group,artnum )  in overview (and other methods not)

Glad to know!

-- 
Julien ÉLIE

« J'ai six loyaux serviteurs, ils m'ont appris tout ce que je sais : ils
   s'appellent quoi, pourquoi, quand, comment, où et qui. » (Rudyard
   Kipling)