postfix_txdelta notification using absolute path (was: Recent binding test failures)

["Peter N. Lundblad" <pe...@famlundblad.se>]

On Mon, 20 Feb 2006, Kouhei Sutou wrote:

> It's OK that the last 'svn ci' success. This is same
> behavior in Ruby bindings test suites.
>
> This problem is that notified informations from
> svn_client_commit() have a garbage. In svn-breakage,
> "/svntest/obj-sh/subversion/bindings/swig/ruby/test/wc-tmp/wc/sample2.txt"
> is the garbage.
>
>
> >    1) Failure:
> >  test_move_force(SvnClientTest) [/srv/svntest/svntest/svn/svn_trunk/subversion/bindings/swig/ruby/test/test_client.rb:969]:
> >  <["test/wc-tmp/wc/sample1.txt", "test/wc-tmp/wc/sample2.txt"]> expected but was
> >  <["/svntest/obj-sh/subversion/bindings/swig/ruby/test/wc-tmp/wc/sample2.txt",
> >   "test/wc-tmp/wc/sample1.txt",
> >   "test/wc-tmp/wc/sample2.txt"]>.
>
>
>

Ah, running the commit in gdb, whatching for calls to notify() shed some
light...

The absolute path comes from the commit_post_txdelta notification, i.e.
when the client is sending the delta between unmodified f1 and modified f1
(to modify f2 after the move).  The reason r18320 broke this test is that
before, no text delta was sent.

It's strange that that particular notification uses an absolute path, but
I tried a "normal commit" (one with only modifications) and that exhibits
the same behaviour.  So, I guess the Ruby test suite has to deal with this
in other places, no?

Actually, looking at the commit code in libsvn_client, the problem seems
to be a simple omission of getting rid of the common path prefix for the
commit when notifying about postfix_txdelta.  Anyone objects to me just
fixing this inconsistency?

Kouhei, I'll take care of this, so you don't need to do anything to the
test suite.

Thanks,
//Peter

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org