Bug 118362 – [ATK] Leak: more leaks in AccessibilityUIElement

• Status: RESOLVED
• Resolution: FIXED
• Component: Accessibility

[Brian Holt]

In Tools/DumpRenderTree/atk/AccessibilityUIElementAtk.cpp:

Leaks found using the "--leak" option in the Gtk port:

{
   <insert_a_suppression_name_here>
   Memcheck:Leak
   fun:malloc
   fun:__vasprintf_chk
   fun:g_vasprintf
   fun:g_strdup_vprintf
   fun:g_strdup_printf
   fun:_ZN22AccessibilityUIElement8languageEv
   fun:_ZL19getLanguageCallbackPK15OpaqueJSContextP13OpaqueJSValueP14OpaqueJSStringPPKS2_
   fun:_ZN3JSC16JSCallbackObjectINS_20JSDestructibleObjectEE14getStaticValueEPNS_9ExecStateENS_12PropertyNameE
   fun:_ZN3JSC16JSCallbackObjectINS_20JSDestructibleObjectEE18getOwnPropertySlotEPNS_6JSCellEPNS_9ExecStateENS_12PropertyNameERNS_12PropertySlotE
   fun:llint_slow_path_get_by_id
   obj:/home/likewise-open/SERILOCAL/brian.holt/Code/gnome3/WebKit/WebKitBuild/Release/.libs/libjavascriptcoregtk-3.0.so.0.14.1
   fun:_ZN3JSC11Interpreter7executeEPNS_17ProgramExecutableEPNS_9ExecStateEPNS_8JSObjectE
   fun:_ZN3JSC8evaluateEPNS_9ExecStateERKNS_10SourceCodeENS_7JSValueEPS5_
   fun:_ZN7WebCore16ScriptController15evaluateInWorldERKNS_16ScriptSourceCodeEPNS_15DOMWrapperWorldE
   fun:_ZN7WebCore16ScriptController8evaluateERKNS_16ScriptSourceCodeE
   fun:_ZN7WebCore13ScriptElement13executeScriptERKNS_16ScriptSourceCodeE
   fun:_ZN7WebCore13ScriptElement13prepareScriptERKN3WTF12TextPositionENS0_17LegacyTypeSupportE
   fun:_ZN7WebCore16HTMLScriptRunner9runScriptEPNS_7ElementERKN3WTF12TextPositionE
   fun:_ZN7WebCore16HTMLScriptRunner7executeEN3WTF10PassRefPtrINS_7ElementEEERKNS1_12TextPositionE
   fun:_ZN7WebCore18HTMLDocumentParser30runScriptsForPausedTreeBuilderEv
   fun:_ZN7WebCore18HTMLDocumentParser16canTakeNextTokenENS0_15SynchronousModeERNS_11PumpSessionE
   fun:_ZN7WebCore18HTMLDocumentParser13pumpTokenizerENS0_15SynchronousModeE
}

{
   <insert_a_suppression_name_here>
   Memcheck:Leak
   fun:malloc
   fun:g_malloc
   fun:g_strdup
   fun:_ZL27webkitAccessibleTextGetTextP8_AtkTextii
   fun:_ZL23webkitAccessibleGetNameP10_AtkObject
   fun:_ZN22AccessibilityUIElement5titleEv
   fun:_ZL16getTitleCallbackPK15OpaqueJSContextP13OpaqueJSValueP14OpaqueJSStringPPKS2_
   fun:_ZN3JSC16JSCallbackObjectINS_20JSDestructibleObjectEE14getStaticValueEPNS_9ExecStateENS_12PropertyNameE
   fun:_ZN3JSC16JSCallbackObjectINS_20JSDestructibleObjectEE18getOwnPropertySlotEPNS_6JSCellEPNS_9ExecStateENS_12PropertyNameERNS_12PropertySlotE
   fun:llint_slow_path_get_by_id
   obj:/home/likewise-open/SERILOCAL/brian.holt/Code/gnome3/WebKit/WebKitBuild/Release/.libs/libjavascriptcoregtk-3.0.so.0.14.1
   fun:_ZN3JSC11Interpreter7executeEPNS_14EvalExecutableEPNS_9ExecStateENS_7JSValueEPNS_7JSScopeE
   fun:_ZN3JSC4evalEPNS_9ExecStateE
   fun:llint_slow_path_call_eval
   obj:/home/likewise-open/SERILOCAL/brian.holt/Code/gnome3/WebKit/WebKitBuild/Release/.libs/libjavascriptcoregtk-3.0.so.0.14.1
   fun:_ZN3JSC11Interpreter7executeEPNS_17ProgramExecutableEPNS_9ExecStateEPNS_8JSObjectE
   fun:_ZN3JSC8evaluateEPNS_9ExecStateERKNS_10SourceCodeENS_7JSValueEPS5_
   fun:_ZN7WebCore16ScriptController15evaluateInWorldERKNS_16ScriptSourceCodeEPNS_15DOMWrapperWorldE
   fun:_ZN7WebCore16ScriptController8evaluateERKNS_16ScriptSourceCodeE
   fun:_ZN7WebCore13ScriptElement13executeScriptERKNS_16ScriptSourceCodeE
   fun:_ZN7WebCore13ScriptElement13prepareScriptERKN3WTF12TextPositionENS0_17LegacyTypeSupportE
   fun:_ZN7WebCore16HTMLScriptRunner9runScriptEPNS_7ElementERKN3WTF12TextPositionE
}

[Brian Holt]

Created attachment 206012 [details]
Patch

[EFL EWS Bot]

Comment on attachment 206012 [details]
Patch

Attachment 206012 [details] did not pass efl-ews (efl):
Output: http://webkit-queues.appspot.com/results/1015764

[Brian Holt]

The title one is a bit trickier because 

   fun:g_strdup
   fun:_ZL27webkitAccessibleTextGetTextP8_AtkTextii
   fun:_ZL23webkitAccessibleGetNameP10_AtkObject
   fun:_ZN22AccessibilityUIElement5titleEv

webkitAccessibleGetName(AtkObject*) sometimes returns a char* that is created by g_strdup but it also sometimes returns a value stored in a cache.  Freeing that would be a disaster.  

My feeling at this stage is that the effort required to refactor is not worth the few bytes leaking.

[EFL EWS Bot]

Comment on attachment 206012 [details]
Patch

Attachment 206012 [details] did not pass efl-wk2-ews (efl-wk2):
Output: http://webkit-queues.appspot.com/results/1012992

[kov's GTK+ EWS bot]

Comment on attachment 206012 [details]
Patch

Attachment 206012 [details] did not pass gtk-ews (gtk):
Output: http://webkit-queues.appspot.com/results/1022262

[Chris Dumez]

Comment on attachment 206012 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=206012&action=review

> Tools/DumpRenderTree/atk/AccessibilityUIElementAtk.cpp:439
> +    GOwnPtr<char> axValue(g_strdup_printf("AXLanguage: %s", locale))

missing ;

> Tools/WebKitTestRunner/InjectedBundle/atk/AccessibilityUIElementAtk.cpp:625
> +    GOwnPtr<char> axValue(g_strdup_printf("AXLanguage: %s", locale))

ditto.

[Brian Holt]

Created attachment 206013 [details]
Patch

[Chris Dumez]

Comment on attachment 206013 [details]
Patch

Ok, r=me. Do you want me to commit? You did not set the cq flag.

[Brian Holt]

(In reply to comment #8)
> (From update of attachment 206013 [details])
> Ok, r=me. Do you want me to commit? You did not set the cq flag.

Yes please, I forgot to do that with webkit-patch.

[WebKit Commit Bot]

Comment on attachment 206013 [details]
Patch

Clearing flags on attachment: 206013

Committed r152371: <http://trac.webkit.org/changeset/152371>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.