From: "Ian Holsman" <ianh@apache.org>
Sent: Monday, January 07, 2002 3:04 PM
> Aaron Bannert wrote:
>
> > On Mon, Jan 07, 2002 at 11:25:55AM -0800, Justin Erenkrantz wrote:
> >
> >>On Mon, Jan 07, 2002 at 10:50:58AM -0800, Aaron Bannert wrote:
> >>
> >>>This is not entirely true. It turns out that although the old shmem API
> >>>is prefixed with apr_shmem, it actually prefixes its functions with just
> >>>apr_shm, which means there is a name conflict between the two destroy
> >>>methods. Looks like we'll have to do a little dance to get this in without
> >>>breaking the builds in too many places.
> >>>
> >>If the patches are done RTC, I'd be comfortable doing it in one pass
> >>(just switch all callers and the API at once). That way we have a
> >>fighting chance to not break the build. -- justin
> >>
> >
> > I agree. I think we have enough people working on this that RTC will work.
> >
> > ** Warning to any apps using APR shared memory: apr_shmem.h will soon
> > be removed. It will be replaced with apr/include/apr_shm.h for the
> > base allocation/mapping routines, and apr-util/include/apr_rmm.h for
> > the memory management routines. **
>
> do you think we should TAG .30 before you introduce this?
Such questions belong on new-httpd, so I've copied that list.
Is the current tree a better codebase than .29, or .28 before that?
If so, we do, if not, we will go ahead and introduce the changes.
I believe it should be apr, then scoreboard [which will -not- use rmm], then
auth_digest, which does use the shm+rmm combo. I'll start on the auth_digest
patch tonight so it can all happen at once.
Bill
|