Re: Store Manager list walk patch

From: Adrian Chadd <adrian@dont-contact.us>
Date: Mon, 8 May 2000 13:43:15 +0800

On Sun, May 07, 2000, Duane Wessels wrote:
>
>
> On Sun, 7 May 2000, Adrian Chadd wrote:
>
> >
> > I've got a patch that kills the list walks whenever a store manager operation
> > is done. The idea is that now you call storeClientCopy() and friends
> > with a store_client pointer, which you get when you call storeClientListAdd().
> > The clients list hanging off the memObject is now a dlink list.
> >
> > The idea is to kill storeClientListSearch() entirely. This should make
> > the store manager slightly faster, but I don't think you can measure the
> > difference before I throw in some more patches.
> >
> > The only thing I'm not entirely sure off is peer_digest.c .. I *think*
> > I have it right, can someone running digests test this for me please?
> >
> > storeClientListSearch() exists only for the assertions I've thrown in to
> > sanity check the code. It passes my repeated polygraph runs.
> >
> > The patch can be found at: http://www.creative.net.au/squid/patches/ .
>
>
> -storeUnregister(StoreEntry * e, void *data)
> +storeUnregister(store_client *sc, StoreEntry * e, void *data)
>
> I doubt you need to pass both 'sc' and 'e' since sc->entry points to 'e'.

Its used for debugging and sanity checking purposes right now.
Once its been running for a while, I'll rip out the extraneous
stuff.

Adrian
Received on Sun May 07 2000 - 23:43:43 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:12:25 MST