rlm_redis fixes

[Gabriel Blanchard]

tx (I wrote the module).

I've been running it in production for months now as well and it's been
running fine.

On 13-02-07 12:28 PM, Brian Candler wrote:
> I have been running a 2.2.0 server with a couple of rlm_redis patches for
> several weeks now with no ill effects (processing some tens or hundreds of
> accounting packets per second).
> 
> They are the top two commits in
> https://github.com/candlerb/freeradius-server/tree/candlerb/redis_escaping
> 
> The earlier one,
> https://github.com/candlerb/freeradius-server/commit/fe7ffbe1e016684a3b791fc86a00fdc7d2f7ccd9
> changes the way that redis escaping is done to allow for values containing
> spaces. It does this by tokenising the source string on spaces and then
> calling xlat on each part individually, building an array of separately
> expanded arguments to pass to redis. This is an approach which was suggested
> to me on the list.
> 
> It is however a behaviour change. Given rlm_rediswho with
>     start-insert = "LPUSH %{User-Name} %l"
> and username "foo bar":
> 
> * originally rlm_rediswho would fail
> * as of commit 6e8e5ab2 it would insert "foo=20bar"
> * with this patch it will insert "foo bar"
> 
> The second one,
> https://github.com/candlerb/freeradius-server/commit/bb7b54637812e787ade8bc27dc215ede2f8653e2
> is a bug fix for rlm_rediswho: the TRIM functionality wasn't working,
> because it was checking the value returned by redis after
> redis_finish_query() was called, and hence was always reading zero.
> 
> Do you think these can be merged?
> 
> Thanks,
> 
> Brian.
> -
> List info/subscribe/unsubscribe? See http://www.freeradius.org/list/devel.html
>