REGRESSION (r109022) Safari not placing service data on pasteboard

• Component: HTML Editing
• Status: RESOLVED FIXED

[Enrica Casucci]

This is a regression introduced during one of the changes to move access to the pasteboard out of the WebProcess.
The support for OS X Services requires that all the write operation to the pasteboard are made synchronously.

<rdar://problem/11085756>

[Enrica Casucci]

Created attachment 138642 [details]
Patch

[WebKit Review Bot]

Attachment 138642 [details] did not pass style-queue:

Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WebCore/ChangeLog', u'Source/WebCor..." exit_code: 1
Source/WebKit2/WebProcess/WebPage/WebPage.h:414:  The parameter name "handle" adds no information, so it should be removed.  [readability/parameter_name] [5]
Total errors found: 1 in 11 files


If any of these errors are false positives, please file a bug against check-webkit-style.

[Enrica Casucci]

(In reply to comment #2)
> Attachment 138642 [details] did not pass style-queue:
> 
> Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'Source/WebCore/ChangeLog', u'Source/WebCor..." exit_code: 1
> Source/WebKit2/WebProcess/WebPage/WebPage.h:414:  The parameter name "handle" adds no information, so it should be removed.  [readability/parameter_name] [5]
> Total errors found: 1 in 11 files
> 
> 
> If any of these errors are false positives, please file a bug against check-webkit-style.

Already fixed.

[Jeff Miller]

Comment on attachment 138642 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=138642&action=review

> Source/WebCore/ChangeLog:8
> +        the pasteboard occurr synchronously. This behavior was changed with r109022.

Type - occurr

> Source/WebKit2/ChangeLog:8
> +        the pasteboard occurr synchronously. This behavior was changed with r109022.

Ditto.

[Jeff Miller]

(In reply to comment #4)
> (From update of attachment 138642 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=138642&action=review
> 
> > Source/WebCore/ChangeLog:8
> > +        the pasteboard occurr synchronously. This behavior was changed with r109022.
> 
> Type - occurr

I meant typo, not type. :)

> 
> > Source/WebKit2/ChangeLog:8
> > +        the pasteboard occurr synchronously. This behavior was changed with r109022.
> 
> Ditto.

[Alexey Proskuryakov]

Comment on attachment 138642 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=138642&action=review

> Source/WebCore/editing/mac/EditorMac.mm:33
>  #import "DocumentFragment.h"
> +#import "DOMRangeInternal.h"

This is out of order, style bit should have complained.

> Source/WebCore/editing/mac/EditorMac.mm:50
>  #import "htmlediting.h"
> +#import "WebNSAttributedStringExtras.h"

This too - upper case comes first in ASCII order.

> Source/WebCore/editing/mac/EditorMac.mm:318
> +PassRefPtr<SharedBuffer> Editor::getBufferSelectionForPasteboard(const String& pasteboardType)

Did this code come from WebKit1? Can it be deleted there now?

This most definitely doesn't look like something we want to maintain in two places forever.

> Source/WebKit2/UIProcess/WebPageProxy.h:473
> +    PassRefPtr<WebCore::SharedBuffer> getBufferSelectionForPasteboard(const String& pasteboardType);

I'm not sure if I can parse "buffer selection". Maybe "PasteboardDataFromSelection"? Or "SerializedAttributedStringFromSelection"?

This applies to the whole cascade of functions, of course, not just this proxy wrapper.

> Source/WebKit2/UIProcess/mac/WebPageProxyMac.mm:-263
> -    if (!isValid())
> -        return false;

Why is this not necessary any more?

> Source/WebKit2/UIProcess/mac/WebPageProxyMac.mm:278
> +    return SharedBuffer::create(static_cast<unsigned char *>(sharedMemoryBuffer->data()), size);

Style nit: star goes with the type.

[Enrica Casucci]

(In reply to comment #5)
> (In reply to comment #4)
> > (From update of attachment 138642 [details] [details])
> > View in context: https://bugs.webkit.org/attachment.cgi?id=138642&action=review
> > 
> > > Source/WebCore/ChangeLog:8
> > > +        the pasteboard occurr synchronously. This behavior was changed with r109022.
> > 
> > Type - occurr
> 
> I meant typo, not type. :)
> 
> > 
> > > Source/WebKit2/ChangeLog:8
> > > +        the pasteboard occurr synchronously. This behavior was changed with r109022.
> > 
> > Ditto.

I'll fix it. Thanks! :-)

[Enrica Casucci]

(In reply to comment #6)
> (From update of attachment 138642 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=138642&action=review
> 
> > Source/WebCore/editing/mac/EditorMac.mm:33
> >  #import "DocumentFragment.h"
> > +#import "DOMRangeInternal.h"
> 
> This is out of order, style bit should have complained.
> 
> > Source/WebCore/editing/mac/EditorMac.mm:50
> >  #import "htmlediting.h"
> > +#import "WebNSAttributedStringExtras.h"
> 
> This too - upper case comes first in ASCII order.
Will do.
> 
> > Source/WebCore/editing/mac/EditorMac.mm:318
> > +PassRefPtr<SharedBuffer> Editor::getBufferSelectionForPasteboard(const String& pasteboardType)
> 
> Did this code come from WebKit1? Can it be deleted there now?


> 
> This most definitely doesn't look like something we want to maintain in two places forever.
> 
No, this code comes from PasteboardMac. I'll refactor it, to avoid duplicates.

> > Source/WebKit2/UIProcess/WebPageProxy.h:473
> > +    PassRefPtr<WebCore::SharedBuffer> getBufferSelectionForPasteboard(const String& pasteboardType);
> 
> I'm not sure if I can parse "buffer selection". Maybe "PasteboardDataFromSelection"? Or "SerializedAttributedStringFromSelection"?
I think I'll go for dataFromSelection and stringFromSelection.

> 
> This applies to the whole cascade of functions, of course, not just this proxy wrapper.
> 
> > Source/WebKit2/UIProcess/mac/WebPageProxyMac.mm:-263
> > -    if (!isValid())
> > -        return false;
> 
Removed by accident. Thanks for noticing.
> Why is this not necessary any more?
> 
> > Source/WebKit2/UIProcess/mac/WebPageProxyMac.mm:278
> > +    return SharedBuffer::create(static_cast<unsigned char *>(sharedMemoryBuffer->data()), size);
> 
> Style nit: star goes with the type.
Sure.

[Alexey Proskuryakov]

> I think I'll go for dataFromSelection and stringFromSelection.

I'm still not sure if that would be clear enough later - when there is a SharedData buffer, it would be helpful to have names explain format for the buffer.

> Removed by accident. Thanks for noticing.

Please add it to the new function as well.

[Enrica Casucci]

http://trac.webkit.org/changeset/115145