http://issues.apache.org/SpamAssassin/show_bug.cgi?id=5254





------- Additional Comments From [EMAIL PROTECTED]  2006-12-31 08:02 -------
Created an attachment (id=3803)
 --> (http://issues.apache.org/SpamAssassin/attachment.cgi?id=3803&action=view)
proposed patch for full write version of syswrite for spamd

Here is a proposed patch to spamd.raw that has a syswrite_full_buffer sub.

This definitely needs critiquing. It seems to solve the problem, but given that
I'm not used to non-blocking I/O in perl and it is very late News Year's Eve
here, I want other eyes on this. I based this on the code in
lib/Mail/SpamAssassin/SpamdForkScaling.pm but I guessed that it is not
necessary to have a 1 second wait between retries here the way that it is done
there. Is that correct? Should the perl in the sub be streamlined any?

I did not restrict this to just the SSL case, as syswrite on any socket could
potentially have this problem.

I tested it in Fedora Core 6. It's too late for me to test it in Cygwin before
uploading it. I'll do that later.

I can't test this in Mac OS X on my Intel MacBook. It turns out that there is
something even more fundamental that breaks trying any SSL I/O in spamd on my
MacBook. SSL fails there with the ironic error message "SSL2_READ_INTERNAL:bad
mac decode" which I'll have to open a separate bug for.

So please look over and critique this patch.




------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

Reply via email to