ra-test.c 13 FAIL over v1 HTTP

[Philip Martin <ph...@codematters.co.uk>]

The test ra-test.c:commit_empty_last_change fails when using the v1 HTTP
protocol:

../src/subversion/tests/libsvn_ra/ra-test.c:1655,
../src/subversion/libsvn_ra_serf/commit.c:1536,
../src/subversion/libsvn_ra_serf/commit.c:406,
../src/subversion/libsvn_ra_serf/commit.c:344,
../src/subversion/libsvn_ra_serf/commit.c:284,
../src/subversion/libsvn_ra_serf/util.c:1032,
../src/subversion/libsvn_ra_serf/util.c:981,
../src/subversion/libsvn_ra_serf/util.c:946,
../src/subversion/libsvn_ra_serf/multistatus.c:558: (apr_err=SVN_ERR_FS_CONFLICT)
svn_tests: E160024: resource out of date; try updating
FAIL:  ra-test 13: check how last change applies to empty commit

The test creates 4 revisions but the three revisions r2-4 all use

 editor->open_root(edit_baton, 1, ...)

i.e. they are explicitly based on r1.  This test was added to assert
some behaviour of the FS backends but I don't think the r1 is necessary
to do that. Should SVN_INVALID_REVNUM be passed or would that defeat the
purpose of the test?
-- 
Philip