bug#46764: Extra ">" sails right past XML validator

bug#46764: Extra ">" sails right past XML validator

[積丹尼 Dan Jacobson]

$ cat e.xml
<?xml version="1.0" encoding="utf-8" ?>
<M>></M>
$ emacs e.xml
says at the bottom: (nXML Valid)
$ xmllint e.kml
<?xml version="1.0" encoding="utf-8"?>
<M>&gt;</M>

bug#46764: Extra ">" sails right past XML validator

[Lars Ingebrigtsen]

積丹尼 Dan Jacobson <jidanni@jidanni.org> writes:

> $ cat e.xml
> <?xml version="1.0" encoding="utf-8" ?>
> <M>></M>
> $ emacs e.xml
> says at the bottom: (nXML Valid)

I can confirm that this problem still exists in Emacs 28.

It seems to stem from this bit of code:

(defun xmltok-forward ()
  (setq xmltok-start (point))
  (let* ((case-fold-search nil)
	 (space-count (skip-chars-forward " \t\r\n"))
	 (ch (char-after)))
    (cond ((eq ch ?\<)
	   (cond ((> space-count 0)
		  (setq xmltok-type 'space))
		 (t
		  (forward-char 1)
		  (xmltok-scan-after-lt))))
	  ((eq ch ?\&)
	   (cond ((> space-count 0)
		  (setq xmltok-type 'space))
		 (t
		  (forward-char 1)
		  (xmltok-scan-after-amp 'xmltok-handle-entity))))
	  ((re-search-forward "[<&]\\|\\(]]>\\)" nil t)
	   (cond ((not (match-beginning 1))

So (xmltok-forward) on the ">" will just return `data'.  Is it checking
just < and & for validity on purpose?  Anybody remember what the thought
process might have been here?

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#46764: Extra ">" sails right past XML validator

[Mattias Engdegård]

">" is not a special character at top level in XML; <M>></M> is well-formed.

I agree that it is an easy mistake to make and overlook. Perhaps an optional warning would be helpful?
Note that nxml-mode is carefully written for correctness and performance, which matter because XML is a lot more complex than people think and files can be large. Any tinkering with it has to be done with prudence.

bug#46764: Extra ">" sails right past XML validator

[Lars Ingebrigtsen]

Mattias Engdegård <mattiase@acm.org> writes:

> ">" is not a special character at top level in XML; <M>></M> is well-formed.
>
> I agree that it is an easy mistake to make and overlook. Perhaps an
> optional warning would be helpful?

Well, if it is valid (and it is), then I don't really see how adding an
optional warning here would be all that helpful, either -- it seems
kinda beyond the remit of the validator here to teach XML syntax?

So I'm closing this bug report.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#46764: Extra ">" sails right past XML validator

[Mattias Engdegård]

26 feb. 2021 kl. 10.30 skrev Lars Ingebrigtsen <larsi@gnus.org>:

>> ">" is not a special character at top level in XML; <M>></M> is well-formed.
>> 
>> I agree that it is an easy mistake to make and overlook. Perhaps an
>> optional warning would be helpful?
> 
> Well, if it is valid (and it is), then I don't really see how adding an
> optional warning here would be all that helpful, either -- it seems
> kinda beyond the remit of the validator here to teach XML syntax?

It's useful to prevent mistakes, not just following the standard to the letter. Given that the XML file is in an Emacs buffer there is a fair chance that it was hand-written, and then the extra ">" is likely to be unintended, especially since it can be somewhat hard to spot by a human.

Many other modes do similar things. For example, emacs-lisp-mode warns about useless backslashes in strings even though there is no actual syntax error. Think of it as a useful compiler warning.

> So I'm closing this bug report.

That's fine. Dan can reopen if he thinks nxml-mode really needs improvement.

bug#46764: Extra ">" sails right past XML validator

[積丹尼 Dan Jacobson]

Fine. Better to let the Space Shuttle engines warn about it than catch it
eariler in emacs, valid or not. P.S.,
$ echo '<M>></M>'|xmllint -
<?xml version="1.0"?>
<M>&gt;</M>
So please file a bug against xmllint.

bug#46764: [External] : bug#46764: Extra ">" sails right past XML validator

[Drew Adams]

> Note that nxml-mode is carefully written for correctness and
> performance, which matter because XML is a lot more complex than people
> think and files can be large. Any tinkering with it has to be done with
> prudence.

+1