Re: Virtual Info files and nodes

[Juri Linkov <juri@jurta.org>]

>>>>> - Why check (stringp filename) in Info-virtual-file?  Isn't it an
>>>>> error to call it with something else than a string.
>>>> Really only nil can be expected now,
>>> Can it?  In which circumstance?
>> `filename' can be nil when `Info-virtual-fun' is called with non-nil
>> `nodename'
>
> That part is obvious enough.
>
>> to check for a node name when a file name is irrelevant -
>> its argument value is nil in this case.
>
> But again: is that ever the case?
> I'd think a node name can only be meaningful when accompagnied by a filename.

There are valid calls with filename=nil where nil means the current
Info file: in Info-find-node-2 filename=nil means to not re-read
the current Info file, other places rely on the value nil returned
from Info-find-file when its argument filename is nil since this
function checks for (stringp filename).  So the same check should be
also in Info-virtual-fun because a call to Info-virtual-fun is outside
of branches that check for a non-string filename in Info-find-file
and Info-find-node-2.

>> This is done in the next version below with other fixes including the
>> handling of "dir":
>
> Thanks, looks fairly good now.  I just worry a little bit about binding
> inhibit-read-only around the call to (Info-virtual-call virtual-fun
> filename nodename no-going-back) since that may involve a lot of code
> that accesses other buffers.  Maybe (setq buffer-read-only nil) would be
> a better choice here.

This code was a copy from a branch that reads a static Info file.
But this could be different like:

	  (let ((filename (or filename Info-current-file)))
	    (setq buffer-file-name nil)
	    (setq buffer-read-only nil)
	    (erase-buffer)
	    (setq Info-current-file filename)
	    (Info-virtual-call virtual-fun filename nodename no-going-back)
	    (set-marker Info-tag-table-marker nil)
	    (setq buffer-read-only t)
	    (set-buffer-modified-p nil)
	    (set (make-local-variable 'Info-current-node-virtual) t))

-- 
Juri Linkov
http://www.jurta.org/emacs/