On Thu, Mar 25, 2010 at 09:06, Julian Foad <julian.foad_at_wandisco.com> wrote:
> Hyrum K. Wright wrote:
>> On Mar 25, 2010, at 7:06 AM, Philip Martin wrote:
>> > OK. I've just noticed that every caller of svn_wc__db_wclock_set
>> > passes zero for levels_to_lock, perhaps it doesn't need to be a
>> > parameter at all.
>>
>> This is here strictly for backward compat. The access batons still
>> have the notion of 'levels to lock' and we need to ensure that they
>> still function correctly, even in the new world. 'levels to lock'
>> should not be exposed through the wc-ng APIs at all: users either get
>> to lock the entire tree (rooted at some subdir, of course), or none.
>
> I added this note of yours in r927378.
>
> Also I fixed another svn_sqlite__bindf "i" parameter type mismatch or
> two in r927344. (Thought: Maybe we should aim to replace
> svn_sqlite__bindf() or its "i" format with some alternative that's
> harder to get wrong.)
Feel free to suggest something, but bindf() is damned handy. I don't
see moving away from that. And adding a second integer type in the
format strings could simply increase the confusion and binding
problems.
Note that there are svn_sqlite__bind_$TYPE() functions that can be
used, whose prototypes automatically correct their args' width.
Cheers,
-g
Received on 2010-03-25 14:29:18 CET