Skip to Content.
Sympa Menu

shibboleth-dev - Re: [Shib-Dev] Idp-ext-delegation & 2.2 idp

Subject: Shibboleth Developers

List archive

Re: [Shib-Dev] Idp-ext-delegation & 2.2 idp


Chronological Thread 
  • From: Halm Reusser <>
  • To:
  • Subject: Re: [Shib-Dev] Idp-ext-delegation & 2.2 idp
  • Date: Fri, 07 Jan 2011 11:04:44 +0100

Hi Brent,

First, thanks for your fast reply although you came just back from vacation.

On 06.01.11 21:53, Brent Putman wrote:
Double-check that in production you implemented step 3 of the
INSTALL.txt with the modifications to shibboleth.OpensamlConfig bean
in internal.xml. There's 3 provider configs that you have to add
there. The one that's either missing or typoed or whatever would be
the /disco-config.xml one, which has the config for that
MetadataAbstract provider.

Damn. You were absolutely right. I didn't remember adjusting
internal.xml, that's why the distribution one was taken instead having
this file in version control. Wrong assumption. My fault.

Note that internal.xml is not reloadable on-the-fly, so you have to
stop and restart the IdP webapp in order for changes to take effect.

I re-deployed this morning. Successfully.

What's basically happening is that the OpenSAML Configuration can't
lookup the builder for that element. I don't like NPE's showing up in
code, but it's an unfortunate fact that throughout the IdP, we don't
check for null when we lookup builders, marshallers and
unmarshallers. We probably should, but it would entail updating
probably thousands of calls.

Concerning the NPE, this wasn't criticism, just a notice. I'm sure I
have a lot of unchecked dereferences in my code.

I'm also not yet sure what's the 'best' approach for that. In that
concrete case we are lucky that the IdP had a global catch (Throwable t)
{...} block, which did logging and therefore helped to easy find the
cause. (Instead of swallowed by stderr and perhaps undiscoverable).

A possible methodology might be 'Design by contract' utilizing
org.springframework.util.Assert for instance and check arguments in
method calls and object construction (the issue here).

But yes - this needs effort or to explain it with an image:
http://xkcd.com/844/


Nevertheless, this was a configuration mistake from my side. Many thanks
for your fast response.


-Halm



Archive powered by MHonArc 2.6.16.

Top of Page