Re: in info reader leading parentheses parsed twice for references?

[Sergey Poznyakoff]

Hi Patrice,

> My understanding of the calls leading to gathering all the xrefs in a node
> (a similar issue with similar calls certainly arises for menu entries) 
> is along
>  
>  info_xref_item 
>  info_select_reference_this_line
>     ->  info_menu_or_ref_item (.... info_xrefs_of_node)
>  info_xrefs_of_node -> info_xrefs -> 
> info_references_internal(INFO_XREF_LABEL, )

Yes, that's quite correct. 

> In info_references_internal INFO_XREF_LABEL is found, then a following :
> is searched for.  If there is another :, the node name is used directly
> (As a side note, I don't really understood if/how the filename is determined 
> in that case if the note is like *Note (file)node::).

To the best of my knowledge I know of no valid way to generate such a
cross reference.  References generated with a @xref command must end with a
dot or comma, otherwise makeinfo bails out.  References generated with a
@pxref should end with a parenthesis.  Finally, those generated with @ref
always end with a dot because makeinfo itself adds it.

Am I missing something?
  
> In info_parse_node, first the (FILENAME) is determined taking into account 
> nested parentheses.  At the end of this code, the position is set at the 
> last correctly nested closed parentheses (l. 109):

To be strict, it is set *past* that parenthesis, unless it ends the line.

> Then node name is parsed with a skip_node_characters (in search.c) call.
> But in skip_node_characters it seems that nested parentheses beginning 
> the string are again taken into account.  Isn't there something strange 
> here?

Yes, it is strange indeed.  I have scanned quite a few existing info
documents, but have not found a single context where this logic might
be needed.  It might well be a leftover from some previous version...  
FWIW, I guess this needs some more investigation.

> I also don't know excactly how to test for those issues as it would require
> the info reader to print somewhere the results of parsing node names and
> file names of cross refs.

That would be a nice feature.  I'll see if it can be incorporated into
the "dump mode" (which is enabled when the `-o' option is given).

Regards,
Sergey