You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Ben Collins-Sussman <su...@collab.net> on 2005/03/07 22:46:18 UTC

mod_dav_svn input filter

Hey Justin E. --

This is a followup discussion to our code-in-progress to send all 
locktokens in the MERGE request body.

I'm utterly surprised.  The request_rec that your input filter stores 
the doc in turns out to be a *different* request_rec than is available 
to mod_dav_svn later on.

I walked through gdb;  the two request_rec structures have different 
addresses, as do their r->pools.  Yet they both look like they're 
"MERGE" requests when I examine them.

How is this possible?  What the heck is going on?

(That certainly explains why apr_pool_userdata_get() was returning 
NULL.  It's a totally different r->pool!  The table technique would 
fail as well.)

I've attached my patch-in-progress for you to look at.


Re: mod_dav_svn input filter

Posted by Ben Collins-Sussman <su...@collab.net>.
On Mar 7, 2005, at 6:57 PM, Ryan Bloom wrote:

> Ben,
>
> I haven't looked at the code, and I have never looked at this, but
> just as a thought, is there a sub-request created and executed during
> processing?  That is really what this sounds like.  You can determine
> this by looking at r->main to find the original request, which is
> where the input_filter will always store data.
>

The only subrequest I'm aware of is the one that mod_dav_svn explicitly 
makes for read-authorization checks, which is a GET request.  I'll take 
a look, though.


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: mod_dav_svn input filter

Posted by Ryan Bloom <rb...@gmail.com>.
Ben,

I haven't looked at the code, and I have never looked at this, but
just as a thought, is there a sub-request created and executed during
processing?  That is really what this sounds like.  You can determine
this by looking at r->main to find the original request, which is
where the input_filter will always store data.

Ryan


On Mon, 7 Mar 2005 16:46:18 -0600, Ben Collins-Sussman
<su...@collab.net> wrote:
> Hey Justin E. --
> 
> This is a followup discussion to our code-in-progress to send all
> locktokens in the MERGE request body.
> 
> I'm utterly surprised.  The request_rec that your input filter stores
> the doc in turns out to be a *different* request_rec than is available
> to mod_dav_svn later on.
> 
> I walked through gdb;  the two request_rec structures have different
> addresses, as do their r->pools.  Yet they both look like they're
> "MERGE" requests when I examine them.
> 
> How is this possible?  What the heck is going on?
> 
> (That certainly explains why apr_pool_userdata_get() was returning
> NULL.  It's a totally different r->pool!  The table technique would
> fail as well.)
> 
> I've attached my patch-in-progress for you to look at.
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
> For additional commands, e-mail: dev-help@subversion.tigris.org
> 
> 
> 


-- 
Ryan Bloom
rbb@apache.org
rbb@rkbloom.net
rbloom@gmail.com

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: mod_dav_svn input filter

Posted by Justin Erenkrantz <ju...@erenkrantz.com>.
--On Monday, March 7, 2005 4:46 PM -0600 Ben Collins-Sussman 
<su...@collab.net> wrote:

> I'm utterly surprised.  The request_rec that your input filter stores the
> doc in turns out to be a *different* request_rec than is available to
> mod_dav_svn later on.

As Ryan guessed, the source structure passed to dav_svn_merge is a sub-req 
produced by dav_lookup_uri and dav_get_resource.  You want to use target 
instead.

Modified patch attached.  -- justin