Re: sendfile in darwin

[William A. Rowe, Jr.]

Jim Jagielski wrote:

On May 13, 2008, at 4:28 PM, William A. Rowe, Jr. wrote:

Jim Jagielski wrote:

I wonder if you are seeing timing issues rather than "real" errors...

are client messages.  However, this could be either 1. timing errors,
or 2. unusual EINTR or other signal noise.

I also wonder if it's cpu based... I'm Intel and you appear
to be PPC.

I more suspect CPU timings; this is a very old 1Ghz box.  Yours is much more
speedy and less prone to actually timing out, I imagine.

Also, if you do these tests back to back, you may also sometimes
see something like:

apr_socket_bind()->48/Address already in use
problem with data read (byte 12495 of trl 3):
read `0' (0x30) from client; expected `@'
apr_socket_recv()->54/Connection reset by peer (expected APR_EOF)

Not seen that, except starting two instances of server ;-)

meaning that the previous server didn't go away fast enough
for the "next" server to startup (if you do a lot of these
tests back to back)... These hand tests seem quite sensitive
to timing.

That's for sure.

Speaking of timing, you can find the truss results of my test in

people.apache.org:~wrowe/sf_dtruss.zip

where the problem is far from obvious :-/

Bill