three questions about ASDF

I have three questions about ASDF.

First, what is a good way to stop ASDF from suppressing the output
describing full WARNINGs which result when it calls SBCL's
COMPILE-FILE function?
    $ cd /tmp
    $ mkdir ./foo
    $ cd ./foo
    $ echo > foo.asd '(defsystem "foo" :components ((:file "foo")))'
    $ echo > foo.lisp '(let ((xxx 1)) (print (- xxxx 1)))'
    $ sbcl --sysinit /dev/null --userinit /dev/null 
    This is SBCL 0.9.13.21, an implementation of ANSI Common Lisp.
  ...
    * (require 'asdf)
    ("ASDF")
    * (asdf:operate 'asdf:load-op 'foo)
  ...
    ; caught STYLE-WARNING:
    ;   The variable XXX is defined but never used.
  ...
    WARNING:
       COMPILE-FILE warned while performing #<COMPILE-OP NIL {AD0B7C1}> on
       #<CL-SOURCE-FILE "foo" {ABE4729}>.
I see STYLE-WARNING output, but not full WARNING output: SBCL signals a
full WARNING that XXXX is undefined, but no corresponding text appears
on my screen, except the summary "COMPILE-FILE warned ...". I have been
playing with expressing the dependencies in a 1KLOC project in ASDF,
and I've made enough coding errors corresponding to full WARNINGs that 
falling back to 
    * (compile-file "foo")
  ...
    ; caught WARNING:
    ;   undefined variable: XXXX
  ...
in order to see the printed form of the full WARNING condition has
become tiresome. Is there a better way?

Second, if I wanted to experiment with the bleeding edge ASDF by using
it to override whatever happens to be bundled with my SBCL
installation, what is "the" (or at least "a") preferred way? Something
like
  (load (compile-file "/usr/local/src/asdf-bleeding-edge/asdf"))
in my .sbclrc?

Third, what is the master upstream version of the ASDF documentation?
I found errours in the version of the documentation frozen in
sbcl/contrib/asdf/ when I was investigating ASDF's errour reporting,
and I'd consider submitting patches, but the documentation in 
"cvs co asdf" from SourceForge looks different, and it isn't 100%
clear to me that even that is the master upstream version of the
documentation, and I don't want to try to fix something which has
already been fixed independently.

-- 
William Harold Newman <[email protected]>
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C  B9 25 FB EE E0 C3 E5 7C
Ubi saeva indignatio ulterius cor lacerare nequit. -- Jonathan Swift's epitaph