On Thu, 2005-08-18 at 14:26, Peter N. Lundblad wrote:
> On Wed, 17 Aug 2005, Madan U Sreenivasan wrote:
[snip]
> > More importantly, the same set of changes are needed every time we
> > need to change the svn_commit_callback_t prototype. am not comfortable
> > with this. shouldnt it be easy to change the callback to accomodate any
> > new info we might need to pass on to the client in future? Any comments?
> >
> >
> This is a good point. In some other places, (notification and status),
> we use a struct as a parameter instead of a collection of parameters. I
> think this is the way to go for callbacks because structs can be extended.
>
> So, why not do the same here?
>
> I propose moving svn_client_commit_info2_t to svn_types.h and calling it
> svn_commit_info_t (you don't need to rev any APIs since that's new). Then
hmmm, replace all usage of svn_client_commit_info2_t to
svn_commit_info_t too? ah, there goes my step 2 down the drain. :(
> rev svn_commit_callback_t to take a const svn_commit_info_t * (and a
> baton, and why not a scratch pool while we are here (see issue #1881)). I
good idea overall. Thx.
I dont know the scratch pool funda. Do you mean that the callback should
take a pool as a parameter?
> know that this will mean reworking some of your previous patches, but I
> think that's the cleanest solution.
sadly, yes, but its okay if we can get some long term leverage because
of this.
>
> We all love API revving, don't we?
really? ;)
Regards,
Maddy.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Aug 19 12:49:05 2005