Adrian Chadd wrote:
> Yup, ditto from here. I'm bughunting a UFS rebuild-from-swaplog bug
> where an object is set to be released but it doesnt come out of the
> replacement policy (grr), so when that is fixed I'll make sure the
> heap replacement code works with UFS.
That might be an oldie. Found a bug like that in 2.2.STABLE5 where
cancelled objects never got released. See my "Squid-2.3.DEVEL2: Release
unused store entries during rebuild" patch.
> diskd will eventually be the async disk IO mechanism squid uses.
On that I am not sure we agree. The design of diskd is basically plauged
by the same issues as async-io (to small granularity of the operations,
not very portable, not performing the best on all platforms where it
builds).
> --enable-diskd and --disable-diskd shouldn't be applicable in
> modio, I'm not sure what is the story in henrik's devel tree.
Neither am I ;-) Well, this area of the code looks as it does in modio
currently.
> the maxobjsize is used as the initial hint a to where to place the
> object. COSS requires it, and I wouldn't wind it above 100k right now.
Is there a explanation of COSS somewhere apart from the actual code?
/Henrik
Received on Sat Apr 22 2000 - 17:35:29 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:12:23 MST