A gotcha that has bitten me quite a few times - when you try to mass-
assign a protected attribute, it fails "silently" (but appears in the
debug log). There is some useful discussion about this subject here:
Perhaps this protected attribute assignment error is worth revisiting
with the addition of the extremely handy rescue_from additions that
have made their way into core?
Of course, I'm getting better about remembering to add attributes via
attr_accessible after being bitten by this one a few times, but
perhaps others have been confounded by this gotcha as well?
I like the idea of assignment to a protected attribute being an
exception. I kinda grok the link to ARes, but it seems that the
change in AR is overly "practical" and doesn't pass the sniff test:
assigning to a protected attribute looks like an exception, it can be
severe (security-wise) and it used to be an exception.
The silent dropping of values bugs me, but in this case I think the
cure is worse than the disease. When we had it enabled previously all
of my exception trackers were spammed with dozens of random junk
coming from adventurous users or broken spam bots.
The current behaviour doesn't have any security related downsides, and
it's just being slightly postel-friendly in the way it behaves.
We could add a hook to make it easier for plugins to handle this
situation, but at present I think it's just a little too annoying for
enabling it by default.
The silent dropping of values bugs me, but in this case I think the
cure is worse than the disease. When we had it enabled previously all
of my exception trackers were spammed with dozens of random junk
coming from adventurous users or broken spam bots.
I would love it if this would only raise an exception in
non-production environments.
Yeah, I'm not sure this "issue" warrants an exception either. Still,
debugging problems that arise due to about attempts to assign
protected (and perhaps read-only) attributes seems unnecessarily
difficult to me at present. I think I've learned my lesson by now, but
perhaps this is a good target for framework polishing? I'm happy to
work on a patch if there's agreement on a smart way to approach the
problem.
Yeah, I'm not sure this "issue" warrants an exception either. Still,
debugging problems that arise due to about attempts to assign
protected (and perhaps read-only) attributes seems unnecessarily
difficult to me at present. I think I've learned my lesson by now, but
perhaps this is a good target for framework polishing? I'm happy to
work on a patch if there's agreement on a smart way to approach the
problem.
I think the first step might be to prise apart that code so that the
handling of protected attributes is handled in a single method like
def handle_unprotected_attribute_assignment(*attributes)
logger.something
end
You could then override that method with a plugin to aid with
debugging during dev and test.