notes on applying a merge changeset

[Denys Duchier]

John A Meinel <john at arbash-meinel.com> writes:

>> There is currently code that attempts to preserve uncommited changes
>> in metadata (the exec flag).  I think that all uncommited changes
>> should be "preserved".  In other words: we should be merging into a
>> "working tree", not merely into a pristine checkout of a committed
>> revision.
>
> Where are we *not* merging into the working tree?

you are not, when (at merge time) modifications in the working tree
are considered illegal.

> I think that merging into a modified tree should be severely
> discouraged. It should be possible to do so with something like the
> current 'merge --force'. But the library should be capable of detecting
> that there are modifications before it starts trying to make any changes.

I suspect you failed to appreciate the point I was making about the
distinction between UI and library API.  The UI should (probably by
default) keep the user from making stupid mistakes.  The library API
should empower the programmer (it should certainly not make things
arbitrarily impossible).  It is fine with me if --force is required in
the UI.  It is not fine with me if the library API makes it impossible
to merge into a working tree with uncommitted changes.

Cheers,

--Denys