Bug 158612 – WTF_CrossThreadTask.Basic fails in all non mac ports

• Component: Web Template Framework
• Status: RESOLVED FIXED

[Carlos Garcia Campos]

../../Tools/TestWebKitAPI/Tests/WTF/CrossThreadTask.cpp:105

Value of: LifetimeLogger::takeLogStr().c_str()

  Actual: "default_constructor(<default>-0-0) copy_constructor(<default>-1-0) name_constructor(logger-0-0) isolatedCopy() copy_constructor(logger-1-0) isolatedCopy() copy_constructor(<default>-2-0) isolatedCopy() copy_constructor(<default>-1-0) move_constructor(logger-1-1) move_constructor(<default>-2-1) move_constructor(<default>-1-1) move_constructor(logger-1-2) move_constructor(<default>-2-2) move_constructor(<default>-1-2) destructor(<default>-1-1) destructor(<default>-2-1) destructor(logger-1-1) destructor(<default>-1-0) destructor(<default>-2-0) destructor(logger-1-0) testFunction called destructor(<default>-1-2) destructor(<default>-2-2) destructor(logger-1-2) destructor(logger-0-0) destructor(<default>-1-0) destructor(<default>-0-0) "

Expected: "default_constructor(<default>-0-0) copy_constructor(<default>-1-0) name_constructor(logger-0-0) isolatedCopy() copy_constructor(<default>-1-0) isolatedCopy() copy_constructor(<default>-2-0) isolatedCopy() copy_constructor(logger-1-0) move_constructor(<default>-1-1) move_constructor(<default>-2-1) move_constructor(logger-1-1) destructor(logger-1-0) destructor(<default>-2-0) destructor(<default>-1-0) move_constructor(<default>-1-2) move_constructor(<default>-2-2) move_constructor(logger-1-2) destructor(logger-1-1) destructor(<default>-2-1) destructor(<default>-1-1) testFunction called destructor(logger-1-2) destructor(<default>-2-2) destructor(<default>-1-2) destructor(logger-0-0) destructor(<default>-1-0) destructor(<default>-0-0) "

FAIL WTF_CrossThreadTask.Basic

C:\cygwin\home\buildbot\slave\win-release\build\Tools\TestWebKitAPI\Tests\WTF\CrossThreadTask.cpp:105
Value of: LifetimeLogger::takeLogStr().c_str()
  Actual: "default_constructor(<default>-0-0) copy_constructor(<default>-1-0) name_constructor(logger-0-0) isolatedCopy() copy_constructor(logger-1-0) isolatedCopy() copy_constructor(<default>-2-0) isolatedCopy() copy_constructor(<default>-1-0) move_constructor(logger-1-1) move_constructor(<default>-2-1) move_constructor(<default>-1-1) move_constructor(logger-1-2) move_constructor(<default>-2-2) move_constructor(<default>-1-2) move_constructor(logger-1-3) move_constructor(<default>-2-3) move_constructor(<default>-1-3) destructor(<default>-1-2) destructor(<default>-2-2) destructor(logger-1-2) destructor(<default>-1-1) destructor(<default>-2-1) destructor(logger-1-1) destructor(<default>-1-0) destructor(<default>-2-0) destructor(logger-1-0) testFunction called destructor(<default>-1-3) destructor(<default>-2-3) destructor(logger-1-3) destructor(logger-0-0) destructor(<default>-1-0) destructor(<default>-0-0) "
Expected: "default_constructor(<default>-0-0) copy_constructor(<default>-1-0) name_constructor(logger-0-0) isolatedCopy() copy_constructor(<default>-1-0) isolatedCopy() copy_constructor(<default>-2-0) isolatedCopy() copy_constructor(logger-1-0) move_constructor(<default>-1-1) move_constructor(<default>-2-1) move_constructor(logger-1-1) destructor(logger-1-0) destructor(<default>-2-0) destructor(<default>-1-0) move_constructor(<default>-1-2) move_constructor(<default>-2-2) move_constructor(logger-1-2) destructor(logger-1-1) destructor(<default>-2-1) destructor(<default>-1-1) testFunction called destructor(logger-1-2) destructor(<default>-2-2) destructor(<default>-1-2) destructor(logger-0-0) destructor(<default>-1-0) destructor(<default>-0-0) "

[Brady Eidson]

The isolated copy snippets are out of order:

Actual:
...
isolatedCopy() 
copy_constructor(logger-1-0) 
isolatedCopy() 
copy_constructor(<default>-2-0) 
isolatedCopy() 
copy_constructor(<default>-1-0) 
...

Expected:
...
isolatedCopy() 
copy_constructor(<default>-1-0) 
isolatedCopy() 
copy_constructor(<default>-2-0) 
isolatedCopy() 
copy_constructor(logger-1-0) 
...

Which is reasonable, considering evaluation order of function arguments is unspecified.

Same thing with the other out-of-orderness later on.

The test assuming evaluation order is bogus.

I'll explore ways to fix.

[Brady Eidson]

Besides out-of-orderness, the exact number of constructor calls per named object is still the same.

So having a map of names-to-constructor calls will work.

[Brady Eidson]

Created attachment 281011 [details]
Patch

[Alex Christensen]

Comment on attachment 281011 [details]
Patch

cool

[Brady Eidson]

http://trac.webkit.org/changeset/201928