On Feb 13, 2009, at 9:50 PM, Greg Stein wrote:
> On Sat, Feb 14, 2009 at 04:46, Hyrum K. Wright
> <hyrum_wright_at_mail.utexas.edu> wrote:
>> On Feb 13, 2009, at 9:27 PM, Greg Stein wrote:
>>> On Thu, Feb 12, 2009 at 07:17, Hyrum K. Wright <hyrum_at_hyrumwright.org
>>> >
>>> wrote:
>>>> ...
>>>> @@ -2141,28 +2091,12 @@ svn_wc_prop_list(apr_hash_t **props,
>>>> SVN_ERR(svn_wc_adm_retrieve(&adm_access, adm_access,
>>>> svn_path_dirname(path, pool), pool));
>>>>
>>>> - return svn_wc__load_props(NULL, props, NULL, adm_access, path,
>>>> pool);
>>>> -}
>>>> -
>>>> -/* Determine if PROPNAME is contained in the list of space
>>>> separated
>>>> - values STRING. */
>>>> + SVN_ERR(svn_wc__load_props(&base_props, &new_props, NULL,
>>>> adm_access,
>>>> + path, pool));
>>>>
>>>> -static svn_boolean_t
>>>> -string_contains_prop(const char *string, const char *propname)
>>>> -{
>>>> - const char *place = strstr(string, propname);
>>>> - int proplen = strlen(propname);
>>>> + *props = apr_hash_overlay(pool, new_props, base_props);
>>>
>>> I don't understand why the logic changed here. The removal of the
>>> propcaching should not have affected the original
>>> svn_wc__load_props()
>>> call. So why the addition of the overlay here?
>>
>> This isn't about propcaching, this is about getting the BASE props
>> and the
>> WORKING props and combining them. With propcaching, the working
>> props were
>> cached (iiuc) so there wasn't a need to fetch them, which is what's
>> happening here.
>>
>> Or, propcaching is just confusing and this makes things work until
>> we get
>> the props moved into sqlite.
>
> propcaching only told you whether one of three properties was present
> on the node: externals, special, or needs-lock. It had nothing to do
> with the values, and reading those from the files. Thus, you shouldn't
> have had to change this code at all.
>
> Simple logic: note that you did *not* modify svn_wc__load_props(), so
> its behavior is unchanged on fetching property values, so you should
> not have had to compensate in any way.
All I know is that with this section reverted, checkout_tests 9
("checkout file with broken eol style") breaks on the branch. I
haven't fully analyzed why, especially since property handling code is
being rewritten Real Soon Now.
-Hyrum
------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=462&dsMessageId=1155362
Received on 2009-02-14 05:18:21 CET