Aaron, I'm backing out the search optimization just to be sure.
Paul J Stevens wrote:
I've done the checkin already. Nothing major wrt searching so I don't
really expect any problems there. I've skipped optimizing _ic_sort, just
because that was less trivial to fix.
below link shows the gist of what I've done:
http://cvs.dbmail.org/cgi-bin/cvsweb.cgi/dbmail/dbsearch.c.diff?r1=1.15.2.4&r2=1.15.2.5&f=h
wrt bug #156:
I've removed the non-functional sigpipe handling in serverchild.c which
was intercepting mysql's own sigpipe handling triggered by mysql_ping.
Finally now, mysql_ping is acting in accordance with the mysql manual.
I've also moved db_check_connection from static to db.h, and have
implemented it in dbpgsql.c also. This way I could add this call in a
couple of obvious places: just before accepting a connection in
serverchild.c, and before handling incoming commands in the protocol
handlers (imap, pop, lmtp).
Aaron Stone wrote:
Paul J Stevens <[EMAIL PROTECTED]> said:
Found the sigpipe bug I believe.
If the mysql connection disappears for some reason (idle timeout, or
mysql restart), the dbmail process should reconnect. It does so for
normal
queries, but in the case of authentication the mysql connection is used
for escaping the login/password sequences without checking the
connection
to the backend.
I was about to say, "That's stupid! Why don't they do it in the
library?!"
but then I thought of a few reasons why they might. And then I decided
they were all stupid reasons. So, back to stupid.
Solution is to call db_check_connection before calling
mysql_real_escape_string.
This is caused by a change the was checked in early november. So this
was
not caused in any way by last minute checkins.
Yep, that's when I did quoting and escaping.
But it was already in 2.0.1. Weird.
Personally I don't see much value in extending the release process. If
people don't test cvs snapshots, they won't test release candidates.
I disagree. Naming is everything. In order to find this problem, a real
site would have had to test out an RC in a (semi-)production environment.
Nobody would do that with a snapshot, but many more admins might try an
RC.
Anyway, I'll check in the fix for this problem together with some
significant search speedups in a couple of minutes.
It might be better to do only this fix, and then release again so that we
have a version out with no major regressions.
Aaron
_______________________________________________
Dbmail-dev mailing list
Dbmail-dev@dbmail.org
http://twister.fastxs.net/mailman/listinfo/dbmail-dev
--
________________________________________________________________
Paul Stevens [EMAIL PROTECTED]
NET FACILITIES GROUP GPG/PGP: 1024D/11F8CD31
The Netherlands_______________________________________www.nfg.nl