On Mon, Sep 14, 2009 at 9:09 AM, Mark Phippard <markphip_at_gmail.com> wrote:
> On Mon, Sep 14, 2009 at 8:57 AM, Mark Phippard <markphip_at_gmail.com> wrote:
>> Paul, I am going to assume this one is for you as I believe you are
>> working on it.
>> It looks like anywhere we call the svn_client_suggest_merge_sources
>> API from JavaHL we get this error. When Java is running SVN API, the
>> current working directory is unlikely to be set to the working copy
>> folder. It looks like maybe somewhere it is making that assumption?
>> Note that the path mentioned is my trunk working copy, not the one for
>> the tests.
> As an additional bit of information, if I set the current directory to
> a path that is not a working copy before the tests run, then it does
> not get any errors and the tests continue.
> Mark Phippard
This failure is traceable back to r38734 which directly fetched an
entry instead of using an access baton and relied on the error
returned when fetching the entry to detect the root of the working
copy. Problem is this is not a very robust way to detect the WC root.
In your particular case, where the root of the WC is in a directory
that is an immediate child of *another* unlrelated WC, the
call to svn_wc__get_entry_versioned() thinks the parent of the WC root
is actually an unversioned path in the encompassing WC
(/Users/mphippard/work/src-trunk), which, since it is a 1.6.x WC*,
causes SVN_ERR_WC_UNSUPPORTED_FORMAT error to be returned.
Anyhow, I fixed r39308 by using svn_wc_is_wc_root2 to detect the root.
* If /Users/mphippard/work/src-trunk *is* a 1.7 WC then let me know,
because then something else is going on!
Received on 2009-09-14 20:48:39 CEST