SharpSvn test suite

r1581401, r1482436, r1482327

r1581401 | rhuijben | 2014-03-25 11:17:13 -0500 (Tue, 25 Mar 2014)

Really cleanup after ourselves when creating a repository fails. Without this
patch we leave an incomplete repository on Windows (and we return an ugly
error, which is why I found this problem).

* subversion/libsvn_repos/repos.c
  (svn_repos_create): Use a scratch pool to hold the lock when creating a
    repository to allow destroying the pool when deleting the repository
    on a failure. Rename the normal pool to result_pool.

Found by: SharpSvn test suite

r1482436 | rhuijben | 2013-05-14 11:31:33 -0500 (Tue, 14 May 2013)

Revert r1293375, to resolve a regression in an optimized code path of
svn_ra_replay_range() and other functions.

* subversion/libsvn_repos/replay.c
  (path_driver_cb_func): Reintroduce the dummy mode documented to apply
    to svn_ra_replay_range().

This will need a further tweak in the editor v2 code to fix the tests that
break by this change.

Found by: SharpSvn test suite

r1482327 | rhuijben | 2013-05-14 08:29:06 -0500 (Tue, 14 May 2013)

* subversion/libsvn_wc/entries.c
  (insert_actual_node): Avoid segfault that was caused by passing NULL to
    strlen. This was triggered on upgrading from entries when there is an
    ACTUAL node to upgrade but no tree conflict.

Found by: SharpSvn test suite
(as regression against 1.7.x)