On Thu, May 16, 2013 at 11:32 AM, Stefan Sperling <s...@elego.de> wrote: > On Thu, May 16, 2013 at 05:07:08PM +0200, Bert Huijben wrote: >> I think we still do, but how can we determine if an incoming 'delete >> <this-exact-tree>' should apply to a node that is not there? > > It sounds like you are still talking about a merge into a sparse WC? > But that is not the problem, as Paul pointed out. The node to be > deleted is there during the merge into an infinite depth WC. > > Anyway, I believe the shallow merge is a red herring. > > During the merge at infinite depth, the tree to be deleted no the > branch does not match the tree deleted on the trunk: > > $ svn diff ^/A/B/E@6 A_COPY/B/E > Index: A_COPY/B/E/beta > =================================================================== > --- A_COPY/B/E/beta > (file:///home/stsp/svn/svn-trunk/subversion/tests/cmdline/svn-test-work/repositories/merge_tests-139/A/B/E) > (revision 6) > +++ A_COPY/B/E/beta (.../A_COPY/B/E) (working copy) > @@ -1 +1 @@ > -New content > \ No newline at end of file > +This is the file 'beta'. > > So I agree that the tree conflict flagged at the end of Paul's test > is warranted. The merge is not deleting the same tree as it was > deleting on trunk. > > If I tweak the test to not run a shallow merge, as below, the test > fails in the same way. > > $ svn diff > Index: merge_tests.py > =================================================================== > --- merge_tests.py (revision 1483398) > +++ merge_tests.py (working copy) > @@ -19026,8 +19026,8 @@ def merge_to_empty_target_merge_to_infinite_target > sbox.simple_commit() > > # r3 - Set depth of A_COPY to empty, merge the r3 from ^/A. > - svntest.actions.run_and_verify_svn(None, None, [], 'up', > - '--set-depth=empty', A_COPY_path) > + #svntest.actions.run_and_verify_svn(None, None, [], 'up', > + # '--set-depth=empty', A_COPY_path) > svntest.actions.run_and_verify_svn(None, None, [], 'merge', '-c7', > '^/A', A_COPY_path) > sbox.simple_commit()
Doh! /me slaps self on forehead Ok, I'm with you on now. My test's expectations are flawed. But Bob's script is a different beast, there is a real problem there. I will rewrite the test to demonstrate his scenario. -- Paul T. Burba CollabNet, Inc. -- www.collab.net -- Enterprise Cloud Development Skype: ptburba
