Re: [PATCH v2 2/3] push: introduce REJECT_FETCH_FIRST and REJECT_NEEDS_FORCE

Jeff King [off-list ref] writes:

On Mon, Jan 21, 2013 at 10:30:29PM -0800, Junio C Hamano wrote:

quoted

When we push to update an existing ref, if:

 * we do not have the object at the tip of the remote; or
 * the object at the tip of the remote is not a commit; or
 * the object we are pushing is not a commit,

there is no point suggesting to fetch, integrate and push again.

If we do not have the current object at the tip of the remote, we
should tell the user to fetch first and evaluate the situation
before deciding what to do next.

Should we? I know that it is more correct to do so, because we do not
even know for sure that the remote object is a commit, and fetching
_might_ lead to us saying "hey, this is not something that can be
fast-forwarded".

But by far the common case will be that it _is_ a commit, and the right
thing is going to be to pull....
Is the extra hassle in the common case worth it for the off chance that
we might give a more accurate message? Should the "fetch first" message
be some hybrid that covers both cases accurately, but still points the
user towards "git pull" (which will fail anyway if the remote ref is not
a commit)?

I was actually much less happy with "needs force" than this one, as
you have to assume too many things for the message to be a useful
and a safe advise: the user has actually examined the situation and
forcing the push is the right thing to do.  Both old and new objects
exist, so the user _could_ have done so, but did he really check
them, thought about the situation and made the right decision?
Perhaps the attempted push had a typo in the object name it wanted
to update the other end with, and the right thing to do is not to
force but to fix the refspec instead?  "You need --force to perform
this push" was a very counter-productive advice in this case, but I
didn't think of a better wording.

The "fetch first and inspect" was an attempt to reduce the risk of
that "needs force" message that could encourage brainless forced
pushes.  Perhaps if we reword "needs force" to something less risky,
we do not have to be so explicit in "You have to fetch first and
examine".

How about doing this?

For "needs force" cases, we say this instead:

 hint: you cannot update a ref that points at a non-commit object, or
 hint: update a ref to point at a non-commit object, without --force.

Being explicit about "non-commit" twice will catch user's eyes and
cause him to double check that it is not a mistyped LHS of the push
refspec (if he is sending a non-commit) or mistyped RHS (if the ref
is pointing at a non-commit).  If he _is_ trying to push a blob out,
the advice makes it clear what to do next: he does want to force it.

If we did that, then we could loosen the "You should fetch first"
case to say something like this:

 hint: you do not have the object at the tip of the remote ref;
 hint: perhaps you want to pull from there first?

This explicitly denies one of Chris's wish "we shouldn't suggest to
merge something that we may not be able to", but in the "You should
fetch first" case, we cannot fundamentally know if we can merge
until we fetch.  I agree with you that the most common case is that
the unknown object is a commit, and that suggesting to pull is a
good compromise.

Note that you _could_ split the "needs force" case into two, namely,
"cannot replace a non-commit" and "cannot push a non-commit".  You
could even further split them into combinations (e.g. an attempt to
replace an annotated tag with a commit and an attempt to replace a
tree with a commit may be different situations), but I think the
advices we can give to these cases would end up being the same, so I
tend to think it is not worth it.  That is what I meant by "I do not
expect me doing the type-based policy myself" in the concluding
message of the series.

On Wed, Jan 23, 2013 at 08:28:49AM -0800, Junio C Hamano wrote:

How about doing this?

For "needs force" cases, we say this instead:

 hint: you cannot update a ref that points at a non-commit object, or
 hint: update a ref to point at a non-commit object, without --force.

Being explicit about "non-commit" twice will catch user's eyes and
cause him to double check that it is not a mistyped LHS of the push
refspec (if he is sending a non-commit) or mistyped RHS (if the ref
is pointing at a non-commit).  If he _is_ trying to push a blob out,
the advice makes it clear what to do next: he does want to force it.

Yeah, I think that is sensible.

Note that you _could_ split the "needs force" case into two, namely,
"cannot replace a non-commit" and "cannot push a non-commit".  You
could even further split them [...etc...]

I do not think it is worth worrying too much about. This should really
not happen very often, and the user should be able to investigate and
figure out what is going on. I think making the error message extremely
specific is just going to end up making it harder to understand.

If we did that, then we could loosen the "You should fetch first"
case to say something like this:

 hint: you do not have the object at the tip of the remote ref;
 hint: perhaps you want to pull from there first?

Yeah, better. I'll comment on the specific message you used in response
to the patch itself.

This explicitly denies one of Chris's wish "we shouldn't suggest to
merge something that we may not be able to", but in the "You should
fetch first" case, we cannot fundamentally know if we can merge
until we fetch.  I agree with you that the most common case is that
the unknown object is a commit, and that suggesting to pull is a
good compromise.

I thought the wish was more about "we shouldn't suggest to merge
something we _know_ we will not be able to", and you are still handling
that (i.e., the "needs force" case).

-Peff