apr-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Cliff Woolley <cliffwool...@yahoo.com>
Subject Re: Buckets destroy not cleaning up private structures?
Date Sun, 29 Apr 2001 04:28:44 GMT
On Sat, 28 Apr 2001, Roy T. Fielding wrote:

> On Sat, Apr 28, 2001 at 09:12:16PM -0700, rbb@covalent.net wrote:
> > the file is closed.  We need a loop at the end of the core_output_filter
> > that just does the conversion to a heap bucket.
>
> The logic should be that anything in the pipeline with a size greater
> than a minimum write (around 4KB I think -- there is a symbol for it
> somewhere) should be written out to the socket and freed before the
> worker moves on to looking for a next request.  Basically, we only
> want to hold onto data if we don't have enough in the buffer to
> justify writing what we have now.

Right, that's the idea.  Anything under the threshold, though, will get
buffered into a heap bucket and the file/mmap/whatever buckets in the
original stream get destroyed so that r->pool can get safely cleaned up.

(And it's 8KB, I'm 99.9% sure.  AP_MIN_BYTES_TO_WRITE ought to be the
symbol of which you speak.)

--Cliff

--------------------------------------------------------------
   Cliff Woolley
   cliffwoolley@yahoo.com
   Charlottesville, VA



Mime
View raw message