Chinmoy Chakraborty <chinmoy12c@xxxxxxxxx> writes:
> Is it safe to make this assumption?
It was the question I asked, and I didn't see a reason to believe it
is safe.
> I mean to be completely sure of this, one would have to track back
> to all the callers.
Yes. If we audit the callers to make sure istate->repo always
points at the right repository, add missing assignment to
istate->repo as necessary, and add
if (!istate->repo)
BUG("caller of cache_tree_udpate() did not fill istate->repo");
then that would be an improvement. But...
> Should a check like:
>
> if(!istate->repo)
>
> istate->repo = the_repository;
>
> be required?
... if we add such an "if the caller did not set istate->repo,
assume the_repository" code, then the resulting code explicitly
assumes that the istate the caller passed to us without setting
istate->repo belongs to the default repository.
I do not quite see the point of such a change---it is not all that
better than "implicit dependency on the_repository" the patch tries
to address, is it?
