[Release Planning 4.8] Samba 4.8.0

[Douglas Bagnall]

On 10/01/18 12:40, Andrew Bartlett wrote:
> On Wed, 2018-01-10 at 00:17 +0100, Stefan Metzmacher wrote:
>> Am 09.01.2018 um 13:22 schrieb Andreas Schneider via samba-technical:
>>> Reminder: We need a release of libldb before Samba 4.8.0rc1! :-)
>>
>> I know :-)
>>
>> Andrew, do we want to consider this?
>> https://git.samba.org/samba.git/?p=mdw/samba.git;a=log;h=refs/heads/ldb_schema.new
>> for https://bugzilla.samba.org/show_bug.cgi?id=8929
>>
>> (But don't push it yet)
> 
> Does it pass all the tests (a full autobuild)?
> 
> Are there any implications for the O() complexity of operations on very
> large lists?
> 
> Also, if we do this I would like to change repl_meta_data to not use
> IGNORE_SINGLE_VALUE_CHECK on creating a backlink and use the attached
> to check it using binary.  However that might also have O()
> implications. 
> 
> Douglas,
> 
> Can you take a look at this?  You did a lot of work here.

There are definitely bits we want no matter what (e.g. this

@@ -265,56 +283,62 @@ int ldb_msg_find_common_values(struct ldb_context *ldb,
 	}
 
 	values = talloc_array(mem_ctx, struct ldb_val, el->num_values);
 	if (values == NULL) {
 		return LDB_ERR_OPERATIONS_ERROR;
 	}
 	values2 = talloc_array(mem_ctx, struct ldb_val,
 				    el2->num_values);
 	if (values2 == NULL) {
+		TALLOC_FREE(values);
 		return LDB_ERR_OPERATIONS_ERROR;
 	}

). The ldb_msg_find_duplicate_val()/ldb_msg_find_common_values()
functions won't change in big-O terms.

My own work in this direction got this far:

http://git.catalyst.net.nz/gitweb?p=samba.git;a=commitdiff;h=f04f93872de45625a4e55bb63b13e4c69525b24b

and then stopped. I can't remember why. It has been a long 7 months.

regards,
Douglas