ActiveRecord prepared statement problem with has_many through with foreign_key

[Robinson7D]

I'm getting the following message when I attempt to use a has_many through with a defined foreign key:

ActiveRecord::StatementInvalid (PG::ProtocolViolation: ERROR:  bind message supplies 2 parameters, but prepared statement "a136" requires 1
: SELECT "social_feeds".* FROM "social_feeds" INNER JOIN "dashboard_tab_feeds" ON "social_feeds"."id" = "dashboard_tab_feeds"."social_feed_id" WHERE "dashboard_tab_feeds"."social_feed_id" = 447 AND "dashboard_tab_feeds"."dashboard_tab_id" = $1):

The query is simply:

# Where tab has already been found, is a DashboardTab:
tab.social_feeds.to_a 

The models are structured as follows:

class DashboardTab < ActiveRecord::Base
  has_many :dashboard_tab_feeds, foreign_key: :dashboard_tab_id, dependent: :destroy
  has_many :social_feeds, through: :dashboard_tab_feeds
end

class DashboardTabFeed < ActiveRecord::Base
  belongs_to :social_feed
  belongs_to :dashboard_tab
end

class SocialFeed < ActiveRecord::Base
  has_many :dashboard_tab_feeds, foreign_key: :social_feed_id, dependent: :destroy
end

This is a trivial example where the foreign keys do not need to be explicitly stated - Rails magic will take care of that for me in this instance - so I can simply remove the foreign_key specifier on the DashboardTab has_many statement and everything works.

This bug seems to be specific to Rails 4.0.1, does not occur on Rails 4.0.0

[senny]

@Robinson7D can you write an executable test-case to reproduce? You can use this script as a foundation.

[vipulnsward]

@Robinson7D @senny I can't seem to reproduce this on master or 4.0.1. at https://gist.github.com/vipulnsward/76a5404d2d4a36a2ec21
@Robinson7D is there some specific call sequence?

[senny]

@vipulnsward never mind reading the issue again reveals that it's about the :foreign_key option not a real foreign key. Sorry for the noise.

[Robinson7D]

Seems like it's working for every test I attempt to create. Perhaps there is an issue with our setup on this project. Vipul's test looks as accurate as can be made to me (and essentially mirrors the version I wrote this weekend based off the test script).

Anyway, since the tests are all passing, and I cannot make them fail, I'll close this ticket.

[vipulnsward]

Thanks @Robinson7D .