On Thu, Jan 31, 2013 at 11:50 AM, Paul Burba <ptbu...@gmail.com> wrote: > On Thu, Jan 31, 2013 at 9:12 AM, <rhuij...@apache.org> wrote: >> Author: rhuijben >> Date: Thu Jan 31 14:12:53 2013 >> New Revision: 1440966 >> >> URL: http://svn.apache.org/viewvc?rev=1440966&view=rev >> Log: >> In the merge handling: make the tree conflict detection for directories work >> like that in the update editor: Detect tree conflicts on entering every node, >> but only report conflicts on actual changes. >> >> This simplifies and brings together the obstruction logic, avoids touching >> unrelated nodes just for testing and makes the tree conflicts more reliable. >> >> The next step will be to hook the file code to this same system. >> >> This patch adds a few notifications for unversioned/missing obstructions >> to make these reports more similar to those of tree conflicts. > . > <SNIP> > . >> def at_directory_external(sbox): >> "tree conflict at directory external" > > Only indirectly related to this commit, but one thing I noticed while > looking at it was a change from 1.7 in the way we report merge skips > within subtrees we don't have read access to. > > Say we have full read access to our merge source (^/A), but only > partial read access to our merge target (^/branch-1): > > authz: > [[[ > [/] > * = rw > [/branch-1/C] > * = > ]]] > > Full disclosure: The above is an unusual[1] authz setup. We already > know the corresponding tree structure in the source, so I don't think > this is so much a security concern as it is a signal-to-noise problem > where potentially a *lot* of skip notifications can be generated. > > Assume a merge wants to apply changes from ^/A into the restricted > subtree of the branch: > > 1.8-dev>svn diff --summarize -c3 > A A\C\J\K\L\M\N\O\P\Q\R\S\T\U\V\W\X\Y\Z\nu > A A\C\J\K\L\M\N\O\P\Q\R\S\T\U\V\W\X\Y\Z > A A\C\J\K\L\M\N\O\P\Q\R\S\T\U\V\W\X\Y > A A\C\J\K\L\M\N\O\P\Q\R\S\T\U\V\W\X > A A\C\J\K\L\M\N\O\P\Q\R\S\T\U\V\W > A A\C\J\K\L\M\N\O\P\Q\R\S\T\U\V > A A\C\J\K\L\M\N\O\P\Q\R\S\T\U > A A\C\J\K\L\M\N\O\P\Q\R\S\T > A A\C\J\K\L\M\N\O\P\Q\R\S > A A\C\J\K\L\M\N\O\P\Q\R > A A\C\J\K\L\M\N\O\P\Q > A A\C\J\K\L\M\N\O\P > A A\C\J\K\L\M\N\O > A A\C\J\K\L\M\N > A A\C\J\K\L\M > A A\C\J\K\L > A A\C\J\K > A A\C\J > > In 1.7 we are only told that the root of the missing subtree is skipped: > > 1.7.9-dev>svn merge ^^/A branch-1 -c3 > Skipped missing target: 'branch-1\C' > --- Recording mergeinfo for merge of r3 into 'branch-1': > U branch-1 > Summary of conflicts: > Skipped paths: 1 > > But with 1.8-dev we see the whole source tree structure revealed via > notifications: > > 1.8-dev>svn merge ^^/A branch-1 -c3 > Skipped missing target: 'branch-1\C' > Skipped 'branch-1\C\J\K\L\M\N\O\P\Q\R\S\T\U\V\W\X\Y\Z\nu' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O\P\Q\R\S\T\U\V\W\X\Y\Z' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O\P\Q\R\S\T\U\V\W\X\Y' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O\P\Q\R\S\T\U\V\W\X' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O\P\Q\R\S\T\U\V\W' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O\P\Q\R\S\T\U\V' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O\P\Q\R\S\T\U' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O\P\Q\R\S\T' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O\P\Q\R\S' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O\P\Q\R' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O\P\Q' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O\P' > Skipped missing target: 'branch-1\C\J\K\L\M\N\O' > Skipped missing target: 'branch-1\C\J\K\L\M\N' > Skipped missing target: 'branch-1\C\J\K\L\M' > Skipped missing target: 'branch-1\C\J\K\L' > Skipped missing target: 'branch-1\C\J\K' > Skipped missing target: 'branch-1\C\J' > --- Recording mergeinfo for merge of r3 into 'branch-1': > U branch-1 > Summary of conflicts: > Skipped paths: 19 > > To be clear, this *doesn't* happen when we don't have access to the > corresponding subtree in the source, regardless of the access we have > to the corresponding subtree in the target: > > authz > [[[ > [/branch-1/C] > * = > [/A/C] > * = > [/] > * = rw > ]]] > > 1.8-dev>svn merge ^^/A branch-1 -c3 > Skipped missing target: 'branch-1\C' > --- Recording mergeinfo for merge of r3 into 'branch-1': > U branch-1 > Summary of conflicts: > Skipped paths: 1 > > authz: > [[[ > [/branch-1/C] > * = rw > [/A/C] > * = > [/] > * = rw > ]]] > > 1.8-dev>svn merge ^^/A branch-1 -c3 > Skipped missing target: 'branch-1\C' > --- Recording mergeinfo for merge of r3 into 'branch-1': > U branch-1\C > U branch-1 > Summary of conflicts: > Skipped paths: 1 > > [1] Seriously, if nobody has ever seen a setup like this in the wild > this is probably a non-issue.
And now we have even odder behavior...see issue #4319: http://subversion.tigris.org/issues/show_bug.cgi?id=4319 -- Paul T. Burba CollabNet, Inc. -- www.collab.net -- Enterprise Cloud Development Skype: ptburba
