Reloading a new record results in an inconsistent state

Matt, the scenario does have code smell, but it's broken nonetheless :)

I should clarify that the ID column is not an auto-increment column.

-1

Without the fix, test doesn't fail on Rails master and Rails 2-3 stable.

Looks fixed already, patch does nothing.

Patch doesn't apply anymore because the relevant code was moved from base.rb to persistence.rb, but the problem still exists. I just tried the test on master and it's still broken. Looking through the code also confirms that new_record is not set to false after successful reload, and save checks "new_record?" and then chooses "create" which does an insert.

+1 test does fail on master.

Rewrote Arya's patch so it applies to master. I've attached a patch and I've attached a test.

IMHO, the correct way to fix this would be to not allow reloading of new records at all. After all, the whole idea of reloading is to re-fetch the record from the database and bring it back in sync with the reloaded object. Since a new record is not yet in the database, reloading does not make sense conceptually.

-1 I agree with David, I don't see the use of reloading a new record.

I don't disagree. In fact that solution was stated in my original ticket. The reason I patched it the way I did was because my particular use case needed me to fetch the record I thought was new, but wasn't, from the database. So if I didn't do it that way, I'd have to do something like record.replace(Model.find(record.id)).

Whether or not use you a non-auto-increment PK is another argument, but that case is clearly supported. So in conclusion, it's still inconsistent, and either solution will resolve the inconsistency.

As far as I'm concerned, I'd rather see #reload raise if passed a new_record?

Automatic cleanup of spam.