On Tue, May 3, 2011 at 3:23 PM, Greg Stein <gstein_at_gmail.com> wrote:
> On Tue, May 3, 2011 at 15:35, Hyrum K Wright <hyrum_at_hyrumwright.org> wrote:
>> On Tue, May 3, 2011 at 2:24 PM, Greg Stein <gstein_at_gmail.com> wrote:
>>> On Tue, May 3, 2011 at 15:04, <hwright_at_apache.org> wrote:
>>>> Author: hwright
>>>> Date: Tue May 3 19:04:24 2011
>>>> New Revision: 1099193
>>>>
>>>> URL: http://svn.apache.org/viewvc?rev=1099193&view=rev
>>>> Log:
>>>> For info, don't fake a completely bogus revision for added nodes, just
>>>> report an invalid revision.
>>>>
>>>> In updating the tests, I noticed we report "Resource is not under version
>>>> control" in the XML for nodes with invalid revnums. While this isn't
>>>> strictly true for added nodes, I'll leave that fix for future change.
>>>
>>> Eh? Shouldn't those tests be marked with XFail, rather than BAD OUTPUT?
>>
>> I'm not sure what you mean. The tests currently pass.
>>
>> My point was that our xml generation assumes an invalid revnum means
>> that thing isn't versioned. While added nodes now correctly report an
>> invalid revnum, they are versioned (or at least we've historically
>> treated them as such, even though they haven't yet been committed to
>> the repository).
>>
>> So spitting out "this thing is not versioned" isn't technically
>> correct, since the node is still under Subversion's control, but it's
>> the best we do right now. I claim the problem is orthogonal to the
>> above change, and should be fixed in the command line client.
>
> You updated the xml test to match the bad output. I think the test
> should be marked XFail (and the output change reverted) until we can
> get the cmdline client fixed to provide the proper output for added
> nodes.
Ah, now I understand.
In my opinion, the current output is just as bogus as before (if more
verbose), so there does not exist a regression. In other words, the
test has been PASS'ing for quite some time, even though it reported
these same nodes as rev=0, which is obviously bogus. If it PASS'd
under those conditions, why not now?
Also, while I'm not planning on fixing this long-standing issue with
the command line client imminently, I'd still like the tests to pass
so as to catch other issues which may come up as I muck with info.
-Hyrum
Received on 2011-05-03 22:29:58 CEST