Emacspeak passing a #<buffer> rather than string as URL base

[Jason Riedy <jason@xxxxxxx>]

I'm trying to use Emacspeak's Google Reader interface, but
somewhere its passing a buffer reference for a URL base rather
than a string.  The macro w3m-string-match-url-components errors
out (through the associated function ...-1) because its argument
is not a string.  This occurs while rewriting local fragment
identifiers, e.g. #someid, in the buffer #<*g scratch*>.

The buffer is temporary, so passing a buffer reference rather
than a string makes sense to me.  But I can't see easily how
emacs-w3m is supposed to handle file-less URLs.  Is using the
buffer's name sufficient?  Or should there be an x-buffer:
protocol?

(A caveat: This is using emacs-w3m from CVS and Emacspeak from
SVN on top of the wonderful emacs-snapshot Debian packages.  So
"go use released versions" is an appropriate response.)

Jason, whose only Japanese-speaking friend uses vi...  ;)

================================================================

System Info to help track down your bug:
---------------------------------------
emacs-w3m-version
 => "1.4.257"
emacs-version
 => "GNU Emacs 23.0.60.1 (x86_64-pc-linux-gnu, GTK+ Version 2.12.7)\n of 2008-02-09 on elegiac, modified by Debian"
mule-version
 => "6.0 (HANACHIRUSATO)"
system-type
 => gnu/linux
(featurep 'gtk)
 => t
w3m-version
 => "w3m/0.5.1+cvs-1.968"
w3m-type
 => w3m-m17n
w3m-compile-options
 => ("lang=en" "m17n" "image" "color" "ansi-color" "mouse" "gpm" "menu" "cookie" "ssl" "ssl-verify" "external-uri-loader" "w3mmailer" "nntp" "gopher" "ipv6" "alarm" "mark" "migemo")
w3m-language
 => nil
w3m-command-arguments
 => nil
w3m-command-arguments-alist
 => nil
w3m-command-environment
 => nil
w3m-input-coding-system
 => utf-8
w3m-output-coding-system
 => utf-8
w3m-use-mule-ucs
 => nil