Bug 45550 – [GTK] Placement new / manual destructor invocation should be used on private GObject memory

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKitGTK

[Martin Robinson]

GLib allocates and deallocates GObject private data structs itself. When those structs contain C++ members, their constructors and destructors are not called. This is not only dangerous, it makes RefPtr-type smart pointers much less useful. We can fix this problem by calling placement new on the private data struct during instance initialization and calling the destructor during finalization.

[Martin Robinson]

Created attachment 67200 [details]
Patch making this change for WebKitWebView

[Xan Lopez]

Comment on attachment 67200 [details]
Patch making this change for WebKitWebView


>-    if (priv->tooltipText) {
>-        gtk_tooltip_set_text(tooltip, priv->tooltipText);
>+    if (priv->tooltipText.length() > 0) {

Mmm, should use isEmpty() for consistency sake?

>+        gtk_tooltip_set_text(tooltip, priv->tooltipText.data());
>         return TRUE;
>     }
> 


Looks great otherwise.

[Martin Robinson]

(In reply to comment #2)

Thanks for the review!

> Mmm, should use isEmpty() for consistency sake?

CString doesn't have an isEmpty unfortunately. I could see it being useful though. Perhaps I'll propose that in another patch.

[Martin Robinson]

Comment on attachment 67200 [details]
Patch making this change for WebKitWebView

Clearing flags on attachment: 67200

Committed r67215: <http://trac.webkit.org/changeset/67215>

[Martin Robinson]

All reviewed patches have been landed.  Closing bug.

[WebKit Review Bot]

http://trac.webkit.org/changeset/67215 might have broken Chromium Mac Release
The following changes are on the blame list:
http://trac.webkit.org/changeset/67216
http://trac.webkit.org/changeset/67215
http://trac.webkit.org/changeset/67205
http://trac.webkit.org/changeset/67206
http://trac.webkit.org/changeset/67207