Bug 188933 – [Attachment Support] Dropping and pasting images should insert inline image elements with _WKAttachments

• Component: WebKit Misc.
• Status: RESOLVED FIXED

[Wenson Hsieh]

Dropping and pasting images should insert inline image elements with _WKAttachments.

[Radar WebKit Bug Importer]

<rdar://problem/43699724>

[Wenson Hsieh]

Created attachment 348055 [details]
First pass

[EWS Watchlist]

Comment on attachment 348055 [details]
First pass

Attachment 348055 [details] did not pass mac-wk2-ews (mac-wk2):
Output: https://webkit-queues.webkit.org/results/8977134

New failing tests:
accessibility/smart-invert-reference.html

[EWS Watchlist]

Created attachment 348064 [details]
Archive of layout-test-results from ews106 for mac-sierra-wk2

The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
Bot: ews106  Port: mac-sierra-wk2  Platform: Mac OS X 10.12.6

[Wenson Hsieh]

Created attachment 348086 [details]
Patch

[Darin Adler]

Comment on attachment 348086 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=348086&action=review

> Source/WebCore/editing/markup.cpp:418
> +            // These attributes are only intended for File deserialization, and are removed from the generated attachment
> +            // element after we've deserialized and set its backing File.

The comment here says "are removed", but doesn’t make it clear where the code that removes them is. Related, a comment like this that makes promises for code elsewhere ("are removed") is dangerous unless the code that does the removing also has a comment pointing back here.

> Source/WebCore/editing/markup.cpp:937
> +    // We use a fake body element here to trick the HTML parser to using the InBody insertion mode.

"trick into" is the normal idiom, rather than "trick to" that we are using here

> Source/WebCore/page/DragController.cpp:1383
> +        attachment = downcast<HTMLAttachmentElement>(&element);

Compiler probably gets this right without help, but it’s theoretically more efficient to do the "&" after casting, rather than converting to a pointer first.

    attachment = &downcast<HTMLAttachmentElement>(element);

> Source/WebCore/page/DragController.cpp:1385
> +    if (is<HTMLImageElement>(element))

Maybe use else if here for clarity?

> Source/WebKit/UIProcess/API/Cocoa/APIAttachmentCocoa.mm:48
> +    return UTTypeIsDeclared((CFStringRef)type) || UTTypeIsDynamic((CFStringRef)type);

Please use __bridge in these typecasts.

> Source/WebKit/UIProcess/API/Cocoa/APIAttachmentCocoa.mm:58
> +    auto mimeType = adoptCF(UTTypeCopyPreferredTagWithClass((CFStringRef)contentType, kUTTagClassMIMEType));
> +    return (NSString *)mimeType.autorelease();

There is no reason to use autorelease here. We can turn a CFString into a WTF::String without involving NSString. Should be:

    return adoptCF(UTTypeCopyPreferredTagWithClass((__bridge CFStringRef)contentType, kUTTagClassMIMEType)).get();

> Source/WebKit/UIProcess/API/Cocoa/APIAttachmentCocoa.mm:68
> +    auto utiType = adoptCF(UTTypeCreatePreferredIdentifierForTag(kUTTagClassMIMEType, (CFStringRef)contentType, nil));
> +    return (NSString *)utiType.autorelease();

There is no reason to use autorelease here. We can turn a CFString into a WTF::String without involving NSString. Also should use a __bridge cast and should not use "nil" for a non-Objective-C-object pointer. Should be:

    return adoptCF(UTTypeCreatePreferredIdentifierForTag(kUTTagClassMIMEType, (__bridge CFStringRef)contentType, nullptr));

[Wenson Hsieh]

Comment on attachment 348086 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=348086&action=review

>> Source/WebCore/editing/markup.cpp:418
>> +            // element after we've deserialized and set its backing File.
> 
> The comment here says "are removed", but doesn’t make it clear where the code that removes them is. Related, a comment like this that makes promises for code elsewhere ("are removed") is dangerous unless the code that does the removing also has a comment pointing back here.

Got it — added to this comment describing where we expect to remove these attributes, and also added a comment where we remove these attributes.

>> Source/WebCore/editing/markup.cpp:937
>> +    // We use a fake body element here to trick the HTML parser to using the InBody insertion mode.
> 
> "trick into" is the normal idiom, rather than "trick to" that we are using here

Fixed!

>> Source/WebCore/page/DragController.cpp:1383
>> +        attachment = downcast<HTMLAttachmentElement>(&element);
> 
> Compiler probably gets this right without help, but it’s theoretically more efficient to do the "&" after casting, rather than converting to a pointer first.
> 
>     attachment = &downcast<HTMLAttachmentElement>(element);

Fixed!

>> Source/WebCore/page/DragController.cpp:1385
>> +    if (is<HTMLImageElement>(element))
> 
> Maybe use else if here for clarity?

Ok!

>> Source/WebKit/UIProcess/API/Cocoa/APIAttachmentCocoa.mm:48
>> +    return UTTypeIsDeclared((CFStringRef)type) || UTTypeIsDynamic((CFStringRef)type);
> 
> Please use __bridge in these typecasts.

Fixed!

>> Source/WebKit/UIProcess/API/Cocoa/APIAttachmentCocoa.mm:58
>> +    return (NSString *)mimeType.autorelease();
> 
> There is no reason to use autorelease here. We can turn a CFString into a WTF::String without involving NSString. Should be:
> 
>     return adoptCF(UTTypeCopyPreferredTagWithClass((__bridge CFStringRef)contentType, kUTTagClassMIMEType)).get();

Good point — fixed!

>> Source/WebKit/UIProcess/API/Cocoa/APIAttachmentCocoa.mm:68
>> +    return (NSString *)utiType.autorelease();
> 
> There is no reason to use autorelease here. We can turn a CFString into a WTF::String without involving NSString. Also should use a __bridge cast and should not use "nil" for a non-Objective-C-object pointer. Should be:
> 
>     return adoptCF(UTTypeCreatePreferredIdentifierForTag(kUTTagClassMIMEType, (__bridge CFStringRef)contentType, nullptr));

Fixed! (with a .get() at the end)

[Wenson Hsieh]

Created attachment 348109 [details]
Patch for landing

[WebKit Commit Bot]

Comment on attachment 348109 [details]
Patch for landing

Clearing flags on attachment: 348109

Committed r235343: <https://trac.webkit.org/changeset/235343>