[OSM-dev] Please advise; I really don't get how this can corrupt so badly

[Dave Stubbs]

2009/4/6 Richard Fairhurst <richard at systemed.net>:
> Stefan de Konink wrote:
>
>> Now how is it possible that there was no 415 on that update?
>
> I'm going to pass on this one as I didn't write the Potlatch
> relations code, Dave did - any thoughts?
>

I'm guessing that's here:

http://trac.openstreetmap.org/browser/sites/rails_port/app/controllers/amf_controller.rb#L309

>>>

 # BUG: the following is commented out because it always fails on my
 #  install. I think it's a Rails bug.

 #if !rel.preconditions_ok?
 #  return -2, "Relation preconditions failed"

<<<

As I remember it is was some kind of composite keys issue throwing a
random exception if you tried to check the relation members... this
was before TomH went into overdrive fixing the composite_keys rails
extension to actually work. It must never have been uncommented.

Anyway, Shaun refactored the whole lot for API 0.6. It now atomically
checks relation consistency within the rails models when doing an
update rather than relying on the caller to do it. So this should be
fixed in a couple of weeks a long with everything else :-)

Dave