On Wed, 28 Feb 2001, Jeff Trawick wrote: > For an immortal bucket, the data is typically a non-writeable string. > But we do a free() on it when we get rid of the bucket representing > it. This seems obviously broken > > Why is transient any different in this respect? The bucket code > doesn't know where the data lives (other than knowing that it needs a > meaningful setaside function).
It was a simple oversight. You were 100% correct to change it. We used to free() the apr_bucket_simple struct. But that layer went away, and I forgot to remove the free(). Thanks, Cliff
