[11864] in Athena Bugs
Re: decmips 7.6G: discuss hangs going to GNU-misc meeting
daemon@ATHENA.MIT.EDU (Ken Raeburn)
Mon Mar 21 15:05:24 1994
From: raeburn@cygnus.com (Ken Raeburn)
Date: Mon, 21 Mar 1994 15:05:03 -0500
To: eichin@MIT.EDU ("Mark W. Eichin")
Cc: ckclark@MIT.EDU, bug-discuss@MIT.EDU, bugs@MIT.EDU
In-Reply-To: eichin@MIT.EDU's message of Sun, 20 Mar 94 21:08:58 EST <9403210208.AA16614@rt-11.MIT.EDU>
GNU-misc was recently expunged - a large number of txn's were removed.
This may be related.
I recall there are certain circumstances where a meeting with a large
number of deleted transactions can be very slow to start up in, if the
client is starting from a deleted txn. You indicate that your
last-read was txn 17756:
The current "first" txn is "1", and the second is "18072". This could
be the problem. It is likely that if you let the client sit there, it
will search through the 316 missing entries, and then it will be fine.
Yes, this is precisely the problem. Without protocol extensions, there
isn't a way to ask the server "what's the first transaction number after
####?", so it has to go looking for it. I fixed the most visible parts
of the problem in goto.c in the ss version some time ago:
revision 1.19
date: 1992/08/06 02:58:55; author: raeburn; state: Exp; lines: +51 -17
If "current" transaction has been deleted, print informational messages
while looking for a non-deleted one. Also, special-case the situation of
transactions 2-N deleted/expunged.
However, it's been an even longer time since /mit/discuss/exl was
updated for anything besides the sun4, never mind the /mit/sipb
versions. We should get a newer version into exl and eventually into
the sipb locker for some real exposure...
