It is with great dismay that I introduce this change and beg forgiveness. We will fix this right. But not today.
Created attachment 27925 [details] Fix-up COM/RefCounted dichotomy in Chromium port, v1. WebCore/ChangeLog | 11 +++++++++++ WebCore/page/chromium/AccessibilityObjectWrapper.h | 9 ++++++++- 2 files changed, 19 insertions(+), 1 deletions(-)
Comment on attachment 27925 [details] Fix-up COM/RefCounted dichotomy in Chromium port, v1. Sigh.
Comment on attachment 27925 [details] Fix-up COM/RefCounted dichotomy in Chromium port, v1. >diff --git a/WebCore/page/chromium/AccessibilityObjectWrapper.h b/WebCore/page/chromium/AccessibilityObjectWrapper.h >index 9920e4d..af32567 100644 >--- a/WebCore/page/chromium/AccessibilityObjectWrapper.h >+++ b/WebCore/page/chromium/AccessibilityObjectWrapper.h >@@ -39,7 +39,14 @@ namespace WebCore { > > protected: > AccessibilityObjectWrapper(AccessibilityObject* obj) >- : RefCounted<AccessibilityObjectWrapper>(0), m_object(obj) { } >+ : m_object(obj) >+ { >+ // Default refCount is 1, but the webkit-win version of this is not >+ // RefCounted, so this is not properly created using ::create. >+ // FIXME: Once AccessibleBase no longer implements COM interfaces, >+ // make it not RefCounted. This comment is a bit confusing. The PLATFORM(WIN) AccessibilityObjectWrapper class does inherit from IUnknown, so it does do initialized-to-zero reference counting as well. Perhaps the comment can be: // FIXME: Remove this once our immediate subclass no longer uses COM. Otherwise, looks good. -Darin
Comment on attachment 27925 [details] Fix-up COM/RefCounted dichotomy in Chromium port, v1. (sorry, didn't mean to remove the r+ from eric)
Landed as http://trac.webkit.org/changeset/41197.