[
https://issues.apache.org/jira/browse/SVN-1611?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Branko Čibej resolved SVN-1611.
-------------------------------
Resolution: Fixed
This was fixed a long time ago with the rewrite of the tree comparison code in
the test framework.
> 'actual' status trees misconstructed in regression test framework
> -----------------------------------------------------------------
>
> Key: SVN-1611
> URL: https://issues.apache.org/jira/browse/SVN-1611
> Project: Subversion
> Issue Type: Bug
> Components: tests
> Affects Versions: all
> Reporter: Karl Fogel
> Assignee: Branko Čibej
> Priority: Critical
> Fix For: unscheduled
>
>
> {noformat:nopanel=true}
> Under certain circumstances, the test suite can misconstruct the 'actual' tree
> in an expected<->actual comparison. Specifically, it can build a node as a
> file
> when it should have built it as a directory.
> For example:
> In r7825 (resolving issue #1609) we added a new test in schedule_tests.py,
> 'status_replaced_directory'. That test passes an expected_status tree to
> run_and_verify_status(). Inside run_and_verify_status(), we do
> output, errput = main.run_svn (None, 'status', '-v', '-u', '-q',
> wc_dir_name)
> mytree = tree.build_tree_from_status (output)
> It happens that wc_dir_name is a greek tree in which 'A' has been 'R'eplaced,
> as in:
> $ svn rm wc_dir_name/A
> $ rm -rf wc_dir_name/A
> $ svn ci -m "Commit deletion of A" wc_dir_name
> $ svn mkdir wc_dir_name/A
> The 'output' passed to build_tree_from_status() therefore looks like this:
> $ svn st -v -u -q wc_dir_name
> R 0 ? ? wc_dir_name/A
> 1 1 jrandom wc_dir_name/iota
> 1 1 jrandom wc_dir_name
> Status against revision: 2
> $
> The returned tree ought to include a node for 'A' as a directory, since that's
> what it is:
> $ ls -l wc_dir_name
> total 8
> drwxr-xr-x 3 kfogel kfogel 4096 Nov 21 07:03 A/
> -rw-r--r-- 1 kfogel kfogel 24 Nov 21 07:03 iota
> $ svn info working_copies/schedule_tests-22/A
> Path: working_copies/schedule_tests-22/A
> URL: file:///blah/blah/blah
> Revision: 0
> Node Kind: directory
> Schedule: replace
> $
> But it does not. For some reason, it builds 'A' as a file, resulting in this
> test failure:
> =============================================================
> Expected A and actual A are different!
> =============================================================
> EXPECTED NODE TO BE:
> =============================================================
> * Node name: A
> Path: working_copies/schedule_tests-22/A
> Contents: None
> Properties: {}
> Attributes: {'status': 'R ', 'repos_rev': '2', 'wc_rev': '0'}
> Children: 4
> =============================================================
> ACTUAL NODE FOUND:
> =============================================================
> * Node name: A
> Path: working_copies/schedule_tests-22/A
> Contents: None
> Properties: {}
> Attributes: {'status': 'R ', 'repos_rev': '2', 'wc_rev': '0'}
> Children: is a file.
> Unequal Types: one Node is a file, the other is a directory
> Unequal at node schedule_tests-22
> Unequal at node working_copies
> (Golly, I hope IssueZilla doesn't wrap that in an ugly way...)
> I believe the problem is in tree.py:
> build_tree_from_status()
> ==> calls create_from_path()
> ==> calls add_elements_as_path()
> ==> calls SVNTreeNode() constructor with children=None
> Since children=None, the node starts out life as a file, not a dir.
> I don't see any point along the call chain where someone could indicate that
> this node is a directory, even if they knew it, which they don't. It's not
> given in the status output, and no one's calling os.path.isdir() to discover
> it.
> How in the heck have we gotten this far with such a bug? :-)
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
