https://issues.apache.org/SpamAssassin/show_bug.cgi?id=6942

--- Comment #17 from AXB <[email protected]> ---
(In reply to Mark Martinec from comment #16)
> Two potential issues:
> 
> - the multi_tok_count_change (called by learn) does not atomically
>   increment token's spam and ham counts. If multiple auto-learning
>   child processes were to learn same tokens, we could lose some increments.
> Should be worry or not to bother?
> Perhaps using GETSET with a callback could be used to detect
> (and later fix) possible collisions.

+1 for worry - If I understand this correctly in a flood where several SA/spamd
sessions are learning the same type of spam, a missed "learn" could delay
higher scored bayes hits.
please correct me if mistaken.


> - as already noted in comments by Henrik, dumping a database
>   (sa-learn --backup) uses a KEYS redis command, which prevents
>   other clients from accessing a database while it is running.
>   Moreover, the way keys() method is implemented in the Redis
>   cpan module, it returns all results in memory. This can make
>   creating a dump(backup) or a large database impossible or
>   just disruptive to other SA processes.
> Should be worry or not to bother?

not worry: +1 - tho a msg "Unsupported feature" sa-learn message should show up
and should this be documented in bayes docs

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to