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

[Sergey Poznyakoff]

Patrice Dumas <address@hidden> ha escrit:
 
> The issue here is not about a missing dot, but about `file' in (file) 
> not being interpreted as a target manual.  Unless I missed something,
> the code also implies that there is no need for a dot following the ::
> as the nodename is immediatly xstrdup'ed from the entry->label
> and no dot or anything else is searched for.

That's right.

> Now on the subject of generating such cross references, it was possible,
> in makeinfo in C by having the manual name as part of the node name,
> although this is certainly an error in the manual,

Ah, I see.  FWIW, I was speaking about "valid ways" of generating
cross-references:)  Anyways, I have verified that info in its current
state follows such references correctly, both in xref and in menu.

Regarding your earlier question:

> (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::).

it is determined as a sequence of characters between '(' and ')' taking
into account eventual balanced parentheses (that's info-utils.c:87-106).

>   *note anode: (afile)anode.
> 
> But I found that weird, and we discussed about it with Karl, I believe
> we did this discussion privately :-(, I proposed to have it rendered as
> 
>   *note (afile)anode::

I see.  That looks better, indeed.

> and Karl agreed as we both thought that the info reader would follow
> such a references by interpreting it as a reference to the node anode 
> in the manual file afile.  But it seems that it is not the case.

No, as far as I can tell, it works OK.  Moreover, so does the Emacs info
reader.  So far I have tested it with manually crafted info files.
I'll test with ones produced by texi2any later.  Just in case, I'm
attaching both testcases.  Let me know if I should improve them.

> In texi2any, I conditionalize the change on a customization variable, 
> set it to false in the default case and set it to true in something 
> like 10 years after the change has been pushed to the Info readers to
> be sure that only the new info readers are used where the new
> manuals are installed.

That's very reasonable, but given that both Emacs and info are already
able to cope with this format, it does not seem necessary.

> That'd be nice, thanks.  Maybe, once you've done that, I could start
> some regression tests of the info reader using this -o feature, by 
> comparing the output with a known reference?

Yes, of course.

Regards,
Sergey

axref.info
Description: Binary data

amenu.info
Description: Binary data