Avoid duplicate internal subsets.

[ruricolist]

I added this change to the previous pull request, but it didn't get merged, I assume because I added it late and it was overlooked. It fixes a bug introduced by the previous fix to DTD embedding, which can result in errors due to duplicate internal subsets and breaks ~60 tests in the SAX test suite (it does not affect Klacks).

[scymtym]

I'm not sure I understand what is going on.

This change causes an error to be signaled from the parser before sax:start-internal-subset is called, but the sink, which must be a sink instance for the have-internal-subset call to work, does its own check:

(defmethod sax:start-internal-subset ((sink sink))
  (when (have-internal-subset sink)
    (error "duplicate internal subset"))
  …)

[scymtym]

Why is the ignore-errors necessary? The only method on have-internal-subset is a slot reader, the slot has :initform nil and there doesn't seem to be a `slot-makunbound' call for the slot.

[ruricolist]

Just running the tests, I see that cxml:sax-proxy does not inherit from cxml:sink, and so does not have an internal-subset slot at all, and rune-dom::dom-builder, while it does inherit from cxml:sink, overrides the definition of the slot and does not provide an initform. Of course there might be others.

[ruricolist]

The bug this pull request addresses only arises with XML documents that have both a DTD and an internal subset.

The patch in 731dd98 solves the problem of getting the serializer to print definitions from DTDs as an internal subset by triggering the start-internal-subset and end-internal-subset events when encountering a DTD.

The problem is that a document could actually have both a DTD and an internal subset, in which case start-internal-subset gets called twice and signals an error.

The simplest thing is to avoid calling start-internal-subset twice and thus avoid the error. Alternately, it might (or might not) be worthwhile to dig into the doctype parsing and serializing code and find a better solution to the problem of serializing documents with DTDs.