commit 623bed05fee6be16c260697526083f89a6541cdf
author Adam Sadovsky <asadovsky@gmail.com> Fri Oct 30 11:02:43 2015 -0700
committer Adam Sadovsky <asadovsky@gmail.com> Fri Oct 30 11:02:43 2015 -0700
tree b7f8e21719ba940e0b2b8e7deb5285e4bf2f1e6a
parent e6ee26c725a4379604826ba502fe9f7ad75f4ec1

test: swap V23InitSimple and V23Init

Also:
- renames the "with mounttable" version to V23InitWithMounttable
- switches most tests to use the "no mounttable" version, since
  they weren't actually using the mounttable
- adds "test.Init()" in a few places that were missing it

When I run "jiri go test v.io/v23/... v.io/x/ref/..." on OS X,
there are still a few warnings about accepting incoming network
connections, but far fewer than before.

The advantage of not including the mounttable for tests where
it's not needed are:
- slightly faster, cheaper tests
- fewer goroutines in panic dumps

Note: Matt expressed a concern that with this change, developers
will be less likely to use the mount table in their tests.

After giving this some thought, I don't think it's the case:
- Developers who understand V23Init/V23InitWithMounttable will
  use whatever's appropriate. (If they really want to use
  endpoints, even though it's trivial to start a mount table,
  we can't stop them.)
- Developers who don't understand V23Init/V23InitWithMounttable,
  and don't think about whether they should use a mount table,
  will just copy the behavior of existing tests, which generally
  use endpoints. This was the case before, and will continue to
  be the case.
- Developers who don't understand V23Init/V23InitWithMounttable,
  but realize they want a mount table, previously would have
  either (a) started their own mount table or (b) looked at the
  impl of V23Init and realized it starts a mount table for them.
  Neither of these scenarios change with this new scheme.

If anything, the new scheme is nice in that developers might by
chance come across an invocation of "test.V23InitWithMounttable",
realize from the name that this is what they should've been using
all along in their own tests, and update those tests to use it
instead of test.V23Init.

(By the argument above: if we really want to have a single helper
that always starts a mount table, we should name it
test.V23InitWithMounttable so that developers can discover that
it has this behavior even while copying code that still uses
endpoints.)

Change-Id: I99c7ca90c144799b472f6d1728aca58ab8128550