[Dbix-class] "search on row->has_many fails if row has prefetch join" after 0.08196 -> 0.08206 upgrade

[Peter Rabbitson]

On Fri, Feb 22, 2013 at 04:46:46PM +0000, Chisel wrote:
> At work we had to freeze our DBIC at 0.08196 until the prefetch/namespace
> bug was fixed ... which it was recently.
> 
> Internally we've built the new perl+cpandeps RPM including DBIx::Class
> 0.08206. Things are looking great for the prefetch/namespace fix but
> something has come to light this week that we think is a bug in DBIC (or
> maybe SQLA).
> 
> The co-worker has managed to distill it into the attached stand-alone test
> case, and the output from the test is below.
> 
> Is this a bug, or are we doing 'bad things' that have come to light with
> newer code in DBIC?

This is a blatant regression, I have no idea how it went undetected :(
Bisection is running while writing this..

> I'm not the resident expert on this problem in-house, but can proxy
> questions through the right channels here if necessary.

The test is comprehensive enough, no further questions. It would be 
great if you could convert it to a DBIC-internal test, for example based 
off of: 
https://github.com/dbsrgits/dbix-class/blob/master/t/prefetch/via_search_related.t 
(otherwise one of us got to do it, tests added to the core suite need to 
follow the internal schema)

> Thanks in advance,

Thanks for finding this, fix should materialize in the next days.