Re: [squid-users] squid 2.6-rproxy: assertion store_client.c:211: "sc->cmp_offset == copy_offset"

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Wed, 12 Jun 2002 00:52:43 +0200

On Tuesday 11 June 2002 22:00, Jeffrey D. Wheelhouse wrote:

> Is the use of http_port_list instead of sockaddr_in_list in the
> rproxy branch a new thing or an old thing?

It is a rproxy thing. rproxy cleans up the http/http_port parsing
slightly as there is now a number of options being added that is
shared by both kinds of HTTP ports.

> > I have merged in the current changes from HEAD. Have not tested
> > if it compiles..
>
> It compiles with warnings (after a few small changes) but does not
> link. Most of the problems are with 3 missing functions:
> cbdataLock, cbdataUnlock, and cbdataValid. It looks like these
> maybe changed names, but I'm not sure which direction. Do these
> translate safely to cbdataInternalLock, etc?

Not exacly. There has been a quite big change in how to use cbdata.

From:
  something = somethinggiventous;
  cbdataLock(something);
To:
  something = cbdataReference(somethinggiventous);

From:
  cbdataUnlock(something);
  something = NULL;
To:
  cbdataReferenceDone(something);

cbdataValid is nowdays cbdataReferenceValid().

And there is also a alternative that is a combination of cbdataValid()
and cbdataReferenceDone(). See the Programmers Guide. It has a whole
chapter on cbdata.

If you find something that does not obviously match the above uses,
please ask for clarification.

> Also, the handling of -a appears to differ from the rproxy branch
> to HEAD in a bad way.
>
> main.c:155 of rproxy reads:
> parse_sockaddr_in_list_token(&Config.Sockaddr.http,
> optarg);
>
> Which appears wrong because &Config.Sockaddr.http is of type
> http_port_list** in the rproxy branch, not sockaddr_in_list**.
>
> Is http_port_list something that's new with rproxy? Or is it
> something old no longer in HEAD?

New, replacing the much simpler sockaddr_in_list thingy..

> Is there a tag to pull the sources from just before the rproxy mod
> was started so I can see better what was explicitly added for
> rproxy? I'm sorry, my CVS knowledge is not the best when it comes
> to all these branches.

There is always a current diff available from
http://devel.squid-cache.org/projects.html#rproxy

Shows the differences of the branch relative to what it is currently
based on.

Yes, there is also a CVS tag, but as the branch is now relative to the
current HEAD this tag will give you the same sources as
Squid-2.6.DEVEL.. If you want ot use CVS the actual tag is
Z-rproxy_merge-HEAD. This tag indicates the version of HEAD the
rproxy branch is currently based on. Changes each time rproxy is
updated to match HEAD using a almost automatic procedure. see
http://devel.squid-cahce.org/

> Two minor patches for the rproxy branch are enclosed below.

Thanks!

Applied with just some minor chages.. (the port issue applies to
ThisCache2 as well..)

Regards
Henrik
Received on Tue Jun 11 2002 - 17:09:39 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:15:40 MST