On Wed, 25 Apr 2001 [EMAIL PROTECTED] wrote:
> > Yes, I agree that you can't use a transient bucket (because a transient
> > bucket means data on the stack that must be passed down right away or
> > setaside (ie, copied into a heap bucket). But you *can* use an immortal
> > bucket. Any disagreement with that statement, Ryan?
>
> I haven't reviewed patch itself, but the concepts are correct.
Patches? We don' need no steeking patches! (Sorry, couldn't resist.)
Anyway, I've changed my mind. I didn't realize that the error message was
configurable within the document itself (nice :-). The fact that it is
means that the error_str gets copied into the include_filter_ctx structure
(which is allocated from f->c->pool) when the context is created and
overwritten if the document specifies its own error message. But it's one
spot in memory that's the one used in all cases. That means that we could
using a pool bucket (not an immortal bucket), but that would only gain us
anything if we create a *single* pool bucket and used apr_bucket_copy().
Not worth the complexity, I think. So I guess it's fine the way it is.
Besides, it's the error path. =-)
--Cliff
--------------------------------------------------------------
Cliff Woolley
[EMAIL PROTECTED]
Charlottesville, VA
