Lars Ingebrigtsen <larsi@gnus.org> writes:

> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>
>>> Xref: reader01.eternal-september.org foo.bar:2523 gnu.emacs.help:57603 zot.bar:3242
>
> [...]
>
>> Slowly, slowly, I'm getting this done. I'm still a bit confused here,
>> though. The xref elements look like they're not supposed to have spaces
>> in them, but the existing code does this:
>>
>> (insert "Xref: " sysname " " group ":")
>> (princ article (current-buffer))
>>
>> Which leaves a space between sysname and group.
>
> I'm not quite sure I understand the question?  The sysname is just a
> part of the syntax of the Xref header and isn't used for anything by
> Gnus, as far as I know.  So there has to be a space?  It's certainly not
> part of the group name.

TBH I only just went and read the RFC for this -- something I'd been
trying to avoid!

>> You say the existing xrefs should be left as they are, but the code adds
>> "prefix" to them. Should this be added unconditionally?
>
> Uhm...  I think so?  But I'm not sure.

Looking over the code again, I think it's best to only add if the prefix
isn't already there.

>> Here's the new version of the function, operating on a header struct.
>> Does this look right to you?
>>
>> Thanks,
>> Eric
>>
>> (defun nnvirtual-update-xref-header (header group prefix sysname)
>>   "Add xref to component GROUP to HEADER.
>> Also add a server PREFIX any existing xref lines."
>>   (let ((bits (split-string (mail-header-xref header)
>> 			    nil t "[[:blank:]]"))
>> 	(art-no (mail-header-number header)))
>>     (setq bits
>> 	  (mapcar (lambda (bit)
>> 		    (concat prefix bit))
>> 		  bits))
>>     (setf (mail-header-xref header)
>> 	  (mapconcat #'identity
>> 		     (cons (format "%s %s:%d"
>>                                    sysname group art-no)
>> 			   bits)
>> 		     " "))))
>
> I think so.  The body of the let form is perhaps more easily expressed
> as
>
> (setf (mail-header-xref header)
>       (concat (format "%s %s:%d " sysname group art-no)
> 	      (mapconcat (lambda (bit)
>  			   (concat prefix bit))
> 			 bits " ")))
>
> ?

Sure, this was just my halfway-there muddle.

I've cleaned this branch, squashed it, and am preparing to test for a
while. I'm attaching the full diff in case anyone wants to read it :)

A net removal of 562 lines with, I hope, no change in behavior.