294115 - [FIX]Eliminate the DummyParserRequest

Status: RESOLVED FIXED

(Core :: DOM: HTML Parser, defect, P1)

Description

[Boris Zbarsky [:bzbarsky]]

I think we can replace the DummyParserRequest with blocking onload via the
nsIDocument api I added in bug 293818.  Need to deal with Cancel(), though; I
think expected behavior on pressing the stop button is in fact that nothing else
will be parsed.  So we may want to terminate the parser even when just
STOP_NETWORK is done (right now we do it for STOP_CONTENT).  Then we can nix
this class altogether.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: Proposed patch

I decided to not worry about Cancel().  Right now it only cancels the parser if we're under the event loop; if we're under parsing (e.g. inline script) it doesn't.  The parser will cancel itself anyway once it gets the OnStopRequest from necko, unless it's already buffered up the whole document...

If we decide we do want to cancel, we can add this to nsDocShell::Stop:

        if (mContentViewer) {
            // Terminate the parser, in case it hasn't happened yet
            nsCOMPtr<nsIDocument> doc;
            mContentViewer->GetDocument(getter_AddRefs(doc));
            if (doc) {
                doc->StopDocumentLoad();
            }
        }

Comment 2

[Blake Kaplan (:mrbkap) (inactive)]

Comment on attachment 254093 [details] [diff] [review]
Proposed patch

Seems reasonable.

Comment 3

[Johnny Stenback (:jst)]

Comment on attachment 254093 [details] [diff] [review]
Proposed patch

Yay, less code! Looks good. sr=jst

Comment 4

[Boris Zbarsky [:bzbarsky]]

Attachment: Updated to tip

Comment 5

[Boris Zbarsky [:bzbarsky]]

Fixed.

Comment 6

[Henri Sivonen (:hsivonen) (not reading bugmail until 2024-09-04)]

Does the code added in this bug still serve a purpose considering that nsDocument::BeginLoad() and nsDocument::EndLoad() block and unblock onload?

Comment 7

[Boris Zbarsky [:bzbarsky]]

Good question.  Maybe not.  Are there cases when DropParserAndPerfHint happen after the content-loaded events fire?