Re: svn commit: r1021795 - introducing svn_wc_add_from_disk()

On Wed, Oct 13, 2010 at 07:16, Julian Foad <julian.foad_at_wandisco.com> wrote:
> On Tue, 2010-10-12, julianfoad_at_apache.org wrote:
>> URL: http://svn.apache.org/viewvc?rev=1021795&view=rev
>> Log:
>> Introduce svn_wc_add_from_disk() as a new API to replace one task that was
>> previously performed by svn_wc_add4().  There are no callers yet.
>
> I updated the libsvn_client callers in r1021800, and those appear to be
> fine.

Excellent. Much nicer to have specific APIs, rather than the catch-all
that add() represented.

> When I update the svn_wc_add4() calls in
> entries-compat.c:test_access_baton_like_locking(), the test fails.
> Which is wrong, the test or my code?

I'm unfamiliar with this test, and would need to research it.

> Note that the test calls first svn_wc_add3() and then svn_wc_add4().  In
> svn_wc_add_from_disk(), I intentionally omitted the "legacy locking"
> code that still exists in svn_wc_add4() line 1033, where it says "If
> using the legacy 1.6 interface ... add is expected to lock the new dir."

Ah. Thanks for the pointers. I'll try to investigate what the failure
pattern here is, over the next few days. (gotta get myself dug back
in!)

>...

re: add4. I agree that we should continue to deprecate add3, remove
our transient add4 definition, and go straight to the more specific
APIs.

Result: we still need to provide add3(), so why does the above testing fail...

Cheers,
-g
Received on 2010-10-13 13:47:22 CEST