Bug 199579 – [WPE][GTK] GUniqueOutPtr::release should return a raw pointer

• Component: WebKitGTK
• Status: RESOLVED FIXED

[Michael Catanzaro]

GUniqueOutPtr::release should release to a raw pointer, rather than a GUniquePtr. If a GUniquePtr is desired, it's trivial to construct one from the raw pointer, but all existing callsites under Source/ would rather have a raw pointer. Currently they have to call release().release() to get the raw pointer, which is annoying.

[Michael Catanzaro]

Created attachment 373644 [details]
Patch

[EWS Watchlist]

Thanks for the patch. If this patch contains new public API please make sure it follows the guidelines for new WebKit2 GTK+ API. See http://trac.webkit.org/wiki/WebKitGTK/AddingNewWebKit2API

[EWS Watchlist]

Comment on attachment 373644 [details]
Patch

Attachment 373644 [details] did not pass win-ews (win):
Output: https://webkit-queues.webkit.org/results/12691399

New failing tests:
fast/block/float/float-with-anonymous-previous-sibling.html

[EWS Watchlist]

Created attachment 373654 [details]
Archive of layout-test-results from ews215 for win-future

The attached test failures were seen while running run-webkit-tests on the win-ews.
Bot: ews215  Port: win-future  Platform: CYGWIN_NT-10.0-17763-3.0.5-338.x86_64-x86_64-64bit

[Adrian Perez]

Comment on attachment 373644 [details]
Patch

As somebody who once had to use a GUniqueOutPtr with a GError
and call .release().release() on it, I am all in favor of having
this landed :)

[Michael Catanzaro]

Comment on attachment 373644 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=373644&action=review

> Source/WTF/wtf/glib/GUniquePtr.h:85
>          GUniquePtr<T> ptr(m_ptr);
>          m_ptr = nullptr;
> -        return ptr;
> +        return ptr.release();

I guess it'd be more efficient to do:

T* ptr = m_ptr.release();
m_ptr = nullptr;
return ptr;

to cut out the extra GUniquePtr. I'll update that before landing.

[Adrian Perez]

(In reply to Michael Catanzaro from comment #6)
> Comment on attachment 373644 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=373644&action=review
> 
> > Source/WTF/wtf/glib/GUniquePtr.h:85
> >          GUniquePtr<T> ptr(m_ptr);
> >          m_ptr = nullptr;
> > -        return ptr;
> > +        return ptr.release();
> 
> I guess it'd be more efficient to do:
> 
> T* ptr = m_ptr.release();
> m_ptr = nullptr;
> return ptr;
> 
> to cut out the extra GUniquePtr. I'll update that before landing.

Wouldn't that be premature optimization? To me it looks like the
compiler should be smart enough to expand the GUniquePtr template
inline and optimize the code to end up with the same assembler code
for both versions 🤔 (unless -O0 is used, of course).

[Carlos Garcia Campos]

Comment on attachment 373644 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=373644&action=review

>> Source/WTF/wtf/glib/GUniquePtr.h:85
>> +        return ptr.release();
> 
> I guess it'd be more efficient to do:
> 
> T* ptr = m_ptr.release();
> m_ptr = nullptr;
> return ptr;
> 
> to cut out the extra GUniquePtr. I'll update that before landing.

m_ptr.release() is not valid, since m_ptr is the raw pointer.

I think you could just do:

return std::exchange(m_ptr, nullptr);

[Michael Catanzaro]

(In reply to Carlos Garcia Campos from comment #8)
> m_ptr.release() is not valid, since m_ptr is the raw pointer.
> 
> I think you could just do:
> 
> return std::exchange(m_ptr, nullptr);

Yes!

[Michael Catanzaro]

Created attachment 373764 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 373764 [details]
Patch

Clearing flags on attachment: 373764

Committed r247294: <https://trac.webkit.org/changeset/247294>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.