commit 1ad5d16dfb3a4d10e173a896c5faba853520f729
author dbates@webkit.org <dbates@webkit.org@268f45cc-cd09-0410-ab3c-d52691b4dbfc> Fri Mar 30 21:01:35 2018 +0000
committer dbates@webkit.org <dbates@webkit.org@268f45cc-cd09-0410-ab3c-d52691b4dbfc> Fri Mar 30 21:01:35 2018 +0000
tree 9e849e64411ac87ba8f139c26ab25cdf4b0b4968
parent d07ae4003651326983747db79e77c2fadbcd3fdf

ASSERTION FAILED: ASSERT(!containsImage || MIMETypeRegistry::isSupportedImageResourceMIMEType([resource MIMEType])) in -[NSPasteboard(WebExtras) _web_writePromisedRTFDFromArchive:containsImage:]
https://bugs.webkit.org/show_bug.cgi?id=184161

Reviewed by Per Arne Vollan.

.:

* ManualTests/DragInlinePDFImageDocument.html: Added.
* ManualTests/resources/simple.pdf: Added.

Source/WebCore:

Fixes an assertion failure when quiting an app that uses a Legacy WebKit web view
after dragging-and-dropping a PDF embedded using an HTML image element into the
same web view.

When performing a drag-and-drop of a PDF document image (WebCore::PDFDocumentImage) we create a WebArchive
from the main frame's WebHTMLView and promise AppKit that will provide a Rich Text Format (RTF) document
from this archive if needed. For some reason, on app termination AppKit requests that the WebHTMLView
fullfill its RTF document promise for the WebArchive created at the start of the drag operation. WebKit
expects that the created WebArchive is either for an inline image (e.g. <img>) or an image document that
has a supported image resource MIME type (by querying MIMETypeRegistry::isSupportedImageResourceMIMEType())
and checks for these cases in this order. PDF/PostScript are not listed in the set of supported image
resource MIME types. So, the first check fails and WebKit assumes that the WebArchive was created from
an image document of a supported image resource MIME type. However, the WebArchive was created from a
WebHTMLView and has MIME type text/html. Therefore the assertion fails. We need to add PDF and PostScript
to the set of supported image resource MIME types so that WebKit does not fall back to the WebHTMLView
code path. Historically, PDF and PostScript were in the set supported image resource MIME types. Over time
the set of MIME types for image resouces (images loaded as a document) became identical to the set of MIME
types for images loaded inline (e.g. <img>) and this set omitted the MIME types for PDF and PostScript.

Additionally it is sufficient to implement MIMETypeRegistry::isSupportedImageResourceMIMEType() in terms
of MIMETypeRegistry::isSupportedImageMIMEType() and MIMETypeRegistry::isPDFOrPostScriptMIMEType() instead
of allocating a dedicated HashSet for the supported image resource MIME types (as we currently do).

* dom/DOMImplementation.cpp:
(WebCore::DOMImplementation::createDocument): Assert that PDF is a supported image MIME type before
instantiating an ImageDocument.
* platform/MIMETypeRegistry.cpp:
(WebCore::initializeSupportedImageMIMETypes): Remove unnecessary allocation of a HashSet for the support
image resource MIME types.
(WebCore::MIMETypeRegistry::isSupportedImageResourceMIMEType): Write in terms of MIMETypeRegistry::isSupportedImageMIMEType()
and MIMETypeRegistry::isPDFOrPostScriptMIMEType().
(WebCore::MIMETypeRegistry::getSupportedImageResourceMIMETypes): Deleted.
* platform/MIMETypeRegistry.h:

git-svn-id: http://svn.webkit.org/repository/webkit/trunk@230117 268f45cc-cd09-0410-ab3c-d52691b4dbfc