fourthTier: Clean up AbstractValue

[Filip Pizlo]

Patch forthcoming.

[Filip Pizlo]

Created attachment 203743 [details]
the patch

I'm going to try to come up with a test case for this...

[Filip Pizlo]

(In reply to comment #1)
> Created an attachment (id=203743) [details]
> the patch
> 
> I'm going to try to come up with a test case for this...

I tried but failed.  This requires a very perverse combination of circumstances to become symptomatic.

[Sam Weinig]

Comment on attachment 203743 [details]
the patch

View in context: https://bugs.webkit.org/attachment.cgi?id=203743&action=review

> Source/JavaScriptCore/dfg/DFGAbstractValue.cpp:220
> +    if (!(m_type & SpecCell)) {
> +        ASSERT(m_currentKnownStructure.isClear());
> +        ASSERT(m_futurePossibleStructure.isClear());
> +        ASSERT(!m_arrayModes);
> +    }
> +    
> +    if (isClear())
> +        ASSERT(!m_value);
> +    
> +    if (!!m_value)
> +        ASSERT(mergeSpeculations(m_type, speculationFromValue(m_value)) == m_type);
> +    
> +    // Note that it's possible for a prediction like (Final, []). This really means that
> +    // the value is bottom and that any code that uses the value is unreachable. But
> +    // we don't want to get pedantic about this as it would only increase the computational
> +    // complexity of the code.

You could consider putting this all in a #if !ASSERT_DISABLED

[Filip Pizlo]

Landed in http://trac.webkit.org/changeset/151229