On Wednesday 02 January 2002 12:21 pm, Bill Stoddard wrote:
> > 1)  Fix the apr_bucket_read macro to check if this bucket is the sentinel,
> > and fail appropriately.  If we do this, we will add one if to every bucket_read
> > operation.
> >
> > 2)  Make the sentinel a more complete bucket along with a bucket->type field.
> > This bucket->type will include a no-op for the read function.
> >
> > I personally prefer option 2, but it may be a bit harder to implement.  I still 
>haven't
> > wrapped my head around the ring macros yet.
> >
> > Ryan
> 
> Yea, 2) looks good in concept, but I'm unsure how APR_BUCKET_REMOVE would be 
>implemented
> to special case removing a sentinel bucket.  The latest fix checked into proxy_util 
>should

It is already possible to try to remove a sentinel bucket, although the result
is undefined, and would most likely make things segfault quickly.  We could do
a quick check in that case, because we try to remove buckets far less often than
we try to read from them.  I am just not sure that the statement above affects
how we should implement the read function.

Ryan
______________________________________________________________
Ryan Bloom                              [EMAIL PROTECTED]
Covalent Technologies                   [EMAIL PROTECTED]
--------------------------------------------------------------

Reply via email to