INN 2.4.0 core dumps
rra at stanford.edu
Wed Feb 21 23:31:47 UTC 2001
Russ Allbery <rra at Stanford.EDU> writes:
> Applied. Teaches me to merge code on vacation without testing it. I'm
> now building current for my test server and will try to flush out any
> remaining problems. Sorry about that.
Okay, I've also fixed another warning nit and moved TMRinit in the innfeed
code to be before the first call to a timer. I'm now getting timers like:
Feb 21 15:20:24 headwall.Stanford.EDU innd: ME time 600003 hishave
41843(28399) hiswrite 39189(6226) hissync 199(624) idle 312564(76257)
artclean 2915(6234) artwrite 13660(2057) artctrl 1698(84) artcncl/artctrl
1696(84) hishave/artcncl/artctrl 589(84) hisgrep/artcncl/artctrl 881(73)
artlog/artcncl/artctrl 24(58) hiswrite/artcncl/artctrl 29(11)
hissync/artcncl/artctrl 0(0) artcncl 166(6) hishave/artcncl 39(6)
hisgrep/artcncl 107(6) artlog/artcncl 3(6) hiswrite/artcncl 0(0)
hissync/artcncl 0(0) sitesend 5419(3964) perl 38222(6226) nntpread
36814(96067) artparse 51060(84827) artlog 3012(6258) datamove 24413(98615)
Note that the nested timer stuff is currently happening due to artcncl
processing and artctrl processing; innreport currently won't do the ideal
thing with this. It all looks fairly reasonable, though.
I'm not sure of the best way of remerging those timers, although I want to
rip out all of the old control message processing from INN as soon as the
new controlchan is in, and that will eliminate the artctrl timer entirely
and collapse some of this.
Another thing that may be a good idea is to move the idle timer into the
timer.h code as the first timer, since I think all callers will use it,
and out of the separate lists of timers in innd and innfeed, which will
restore it to its position as the first timer. I think that's more
convenient for a quick scan.
Russ Allbery (rra at stanford.edu) <http://www.eyrie.org/~eagle/>
More information about the inn-workers