CURRENT branch

[Russ Allbery]

Julien ÉLIE <julien at trigofacile.com> writes:

>> Or, even more simply, just set a flag whenever we do any writes to
>> the overview at all and always remap the data in that case, but keep
>> the previous logic for when to remap the index.

> But then, we would also do more than we have to.  If 1-6 are mapped,
> article #7 arrives and I search in the range 3-6, there is no need
> to remap the data file.

Yes, true.

All these fixes don't help the case of out-of-order delivery and nnrpd,
unfortunately, since nnrpd isn't doing the writing and hence won't get
the out-of-order flag set and won't know to remap.  That was the
scenario the code that Jeff removed was originally intended to deal
with, but it also breaks OVSTATICSEARCH.  I think the current tradeoff
(where the article doesn't appear) is probably better than breaking
OVSTATICSEARCH, but it does mean that if we do out-of-order article
delivery while nnrpd has a search open, the client will end up never
seeing that article.

It's likely to be quite rare that this will happen, though.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>

    Please send questions to the list rather than mailing me directly.
     <http://www.eyrie.org/~eagle/faqs/questions.html> explains why.