stress testing of commloops

Hi guys,

I've reached the end of this particular phase of the commloops
code modifications, and I'd like some general feedback.

Basically, I've:

* removed the current range request logic, so for now all
  range requests are treated as uncacheable and passed through
* tried to use a static buffer rather than memalloc'ed buffers when
  providing memory to storeClientCopy()
* removed the seen_offset from storeClientCopy, redoing (most!) of the
  logic which used this for header parsing

My next moves will include turning storeClientCopy() into the exact
opposite of storeAppend, and then looking for a nice and tidy way
to tie all of this evilness into something eventio-friendly.

(most - indicates the netdb and peerdigest code still doesn't "use" the
 new storeClientCopy API, but its not that hard to do if someone feels
 inclined..)

My current todo list is at http://devel.squid-cache.org/commloops/

Right now I'm just looking for some testing (both polygraph and
browsing) and some code feedback before I go forward.
I don't expect this to go into squid-2.5, but it'd be nice to finish
everything on my TODO list above and place it into squid-HEAD once
2.5 has forked, so people get the "hang" of this new API.

Robert - note how this is friendlier to your content modification/te
stuff. :-)

Adrian
Received on Thu Oct 18 2001 - 02:56:09 MDT