[Monotone-devel] disapprove followed by re-propagate?

[Shawn Samuel]

Conceptually, what happens in the following case:

a) propagate branch b (rev bb) -> branch a (currently at rev 1), results in rev 
aa on branch a.
b) disapprove rev aa (results in rev aa1)
c) check in changes to rev bb (on branch b) that fix whatever made bb bad (rev 
bb1)
d) propagate branch b (rev bb1) -> branch a (rev aa2)

Intuitively, it seems that since bb is an ancestor of both heads at
this point, that the reversals done by aa1 will stay in place (for the
set of changes in bb that weren't touched by bb1, call it bb-bb1),
since the bb changes are already in both branches, and the reversals
are the only changes that came later than that (for bb-bb1).

The workflow I'm thinking about is:

1) developer propagates bad changes to trunk
2) changes get disapproved so other devs don't pick it up.
3) developer fixes bad parts of change and re-propagates to trunk

For this to work, it seems to me that I'd have to disapprove aa1, then
propagate bb1 in order to get what I want. What I want, in case it's
not obvious, is what would have happened if the dev had never done the
first propagate and just propagated from branch b(rev aa1) -> branch a(rev 1).

Or is there something else that I should be doing in step 2 of the workflow?

thanks,
shawn