Re: svn commit: r923720 - in /subversion/trunk/subversion/libsvn_client: client.h diff.c repos_diff.c

Greg Stein <gstein_at_gmail.com> writes:

>> +/* Return in *LOCAL_DIR_ABSPATH the absolute path for the directory
>> +   PATH if PATH is a versioned directory.  If PATH is not a versioned
>> +   directory and LENIENT is FALSE then return an error
>> +   SVN_ERR_WC_NOT_WORKING_COPY.  If LENIENT is TRUE then failure to
>> +   find an access baton will not return an error but will set
>> +   *LOCAL_DIR_ABSPATH to NULL instead.
>> +
>> +   This rather odd interface was originally designed around searching
>> +   an access baton set. */
>
> "failure to find an access baton" ?? That certainly doesn't make any
> sense.

Oops. I wrote the doc string and reverted the patch several times,
obviously I got a bit sloppy with the final version.

> The docstring should also mention that if WC_CTX is NULL, then
> nothing will ever be returned (*LOCAL_DIR_ABSPATH is set to NULL).
> IMO, this function shouldn't even be called in that situation. Callers
> should not even be trying functionality which doesn't make sense (ie.
> push the WC_CTX check out).

It's a very weird function now because in 1.6 it was a wrapper around
svn_wc_adm_retrieve and the interface kind of matches that function.
I'll look at whether checking WC_CTX in the caller is better.

-- 
Philip