This is a Linux oddity. If you look at
http://www.ussg.iu.edu/hypermail/linux/kernel/0202.3/0292.html you'll see a
discussion. The code calls setsockopt to suggest the desired value and
then calls getsockopt to see what it was actually set to - I didn't want
you to request 1M and, unbeknownst to you, use 1K. Maybe I should get rid
of the warning. Opinions?
Greg
David Thompson
<[EMAIL PROTECTED]> To:
[email protected]
Sent by: cc:
[EMAIL PROTECTED] Subject: Re: [opendx-users] 4.3
Socket Problem
son.ibm.com
07/22/2003 01:56 PM
Please respond to
opendx-users
Its not actually an error, it is more a warning. But I don't
understand where it could be coming up with the doubling affect. It
is using setsockopt to increase the bufsize and then rereads what the
system actually set with getsockopt and compares the two. Have you
tried setting it lower, say 131071?
The system default for many Linuxes is 65535, but this may be too low
and is exhibited by a dead lock condition when loading some large
networks.
David
>Thanks, David. In /usr/local/dx/bin/dx, around line 464, the script
>already reads as you suggest (below). Do I have the wrong version?
>
>As a test, I commented out the code you show below and dx ran with
>no errors. Then I repeated the test but with DX_SOCKET_BUFSIZE set
>to 500000; then, the error returns with yyy zzz ===> 500000 1000000.
>
>It looks like DX is always obtaining a buffer twice the size it
>wants except when using defaults.
>
>Kent
>
--
.............................................................................
David L. Thompson Visualization and Imagery Solutions,
Inc.
mailto:[EMAIL PROTECTED] 5515 Skyway Drive, Missoula, MT 59804
Phone : (406)756-7472
