On Thu, Feb 14, 2002 at 05:49:59PM -0500, Kevin Pilch-Bisson wrote:
> On Thu, Feb 14, 2002 at 09:17:44PM +0000, Joe Orton wrote:
> > That looks right yes. Can you try turning on neon debugging in
> > session.c, and post the debug log somewhere?
>
> debug log is at http://pilchie.homeip.net/~kevin/co.txt
Thanks!
Of note is that mod_deflate seems to be buffering the entire response in
memory before sending it to the client, and using a fixed-length
response (precomputed Content-Length) rather than streaming it using
chunked transfer-encoding. This comment in the source would indicate
that this is a known problem:
/* XXX FIX: do we need the Content-Size set, or can we stream?
* we should be able to stream */
/* ignore flush buckets for the moment.. we can't stream as we
* need the size ;( */
For the response that fails, the GET for fs-test.c, the server is
reporting a Content-Length which matches the actual size of fs-test.c in
my working copy, 213750, whereas the Content-Length (if used) *should*
be equal to the compressed size, as it is for all the other responses.
After 15 blocks of response reach the client, totalling 38991 bytes, the
connection is closed by the server.
So as far as I can tell here, this is a server problem.
There is one debug log message in mod_deflate, which will appear if you
use "LogLevel debug" in Apache; "Zlib: Compressed %ld to %ld : URL %s" -
it might be useful to turn that on and see what it says for fs-test.c.
Regards,
joe
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:37:08 2006