svndumpfilter matches by node path string containment rather than by node path directory containment

If one attempts to filter out a given path from a repository dump using svndumpfilter, the result is that all path strings beginning with the given path are filtered out. What should happen is that only path nodes equal to or contained by the given node should be filtered out.

Example:

I have a repository containing 2 nodes at the top level:
  Directory "Foo"
  Plain file "FooBar.txt"

When I ask svndumpfilter to remove just the "Foo" node, both the "Foo" and the "FooBar.txt" nodes are removed.

Specifying "Foo/" as the path to remove doesn't help.

I have attached a 695-byte tarball, "svndumpfilter_bug_demo.tar.gz", to demonstrate the problem.

The attached file contains 4 files:
  - repo_dump (the result of "svnadmin dump /my/repo")
  - svndumpfilter_cmd (the svndumpfilter command line I used)
  - svndumpfilter.out (the stderr produced by svndumpfilter)
  - repo_dump_no_Foo (the stdout produced by svndumpfilter)

As you can see from svndumpfilter_cmd, I am attempting to filter out just the "Foo" node. But, as you can see from svndumpfilter.out (and repo_dump_no_Foo), the "FooBar.txt" node is incorrectly filtered out.

The above issue was mentioned in the comment by Max Bowsher dated 2005/9/29 in the unrelated subversion issue #1853 (svndumpfilter forgetting parent dirs).
http://subversion.tigris.org/issues/show_bug.cgi?id=1853

I realize that the subversion manual states that "svndumpfilter takes path filtering quite literally", however the issue that the manual goes on to describe there is an entirely separate issue.

Do you agree that this is a problem? Shall I file an issue for this in the subversion issue tracker?

Thank you.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org