[merge] Ignore write errors when creating hashcache

[Aaron Bentley]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

John Arbash Meinel wrote:
> Also, I don't know how Hg or Git would handle a test like this.

The Git plugin isn't remotely close to being testable with our standard
framework.  It doesn't have any kind of WorkingTree at all.

> Which is why I wasn't positive that we wanted it as a generic working
> tree test. Except it seems like something we *would* want out of all
> working trees.

I think it's better to have a spurious failure than a false success.

I think all trees should guarantee that caching in readonly trees
behaves correctly.  Or if we want to permit certain trees to break, we
should give them specific exemptions.

So if that test fails, we can make it work with a new tree type.  But if
it falsely succeeds, we'll never know.

Worst case, we could do sleep(3) for trees without hashcaches.

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFFiEBS0F+nu1YWqI0RAn7yAJ9X+KhZ5qbzB/oqmYPb3jSELC87EgCghX8D
Crsc1pq6W1PH0rdRl07MoKc=
=br+S
-----END PGP SIGNATURE-----