Bug 203142 – [Clang][Windows] Options.cpp(317,25): error: no matching function for call to 'optionTypeSpecificIndex'

• Component: JavaScriptCore
• Status: RESOLVED FIXED

[Fujii Hironori]

Created attachment 381272 [details]
compilation error

Since r251226 (Bug 202105)
Clang-cl 8 can't compile WinCairo trunk.


> FAILED: Source/JavaScriptCore/CMakeFiles/JavaScriptCore.dir/runtime/Options.cpp.obj 
> C:\PROGRA~1\LLVM\bin\clang-cl.exe  (...) -c ..\..\Source\JavaScriptCore\runtime\Options.cpp
> ..\..\Source\JavaScriptCore\runtime\Options.cpp(294,12): error: variables defined in a constexpr function must be initialized
>     size_t index;
>            ^
> ..\..\Source\JavaScriptCore\runtime\Options.cpp(317,25): error: no matching function for call to 'optionTypeSpecificIndex'
>     FOR_EACH_JSC_OPTION(DEFINE_OPTION_ACCESSORS)
>     ~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~
> ..\..\Source\JavaScriptCore\runtime/OptionsList.h(124,5): note: expanded from macro 'FOR_EACH_JSC_OPTION'
>     v(Bool, useKernTCSM, true, Normal, "Note: this needs to go before other options since they depend on this value.") \
>     ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> ..\..\Source\JavaScriptCore\runtime\Options.cpp(311,49): note: expanded from macro 'DEFINE_OPTION_ACCESSORS'
>         return g_jscConfig.type##type_##Options[optionTypeSpecificIndex<OptionTypeID::type_, OptionID::name_>()]; \
>                                                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> ..\..\Source\JavaScriptCore\runtime\Options.cpp(292,18): note: candidate template ignored: substitution failure [with type = JSC::OptionTypeID::Bool, id = JSC::OptionID::useKernTCSM]
> constexpr size_t optionTypeSpecificIndex()
>                  ^

[Fujii Hironori]

Created attachment 381277 [details]
Patch

[Mark Lam]

Comment on attachment 381277 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=381277&action=review

> Source/JavaScriptCore/ChangeLog:8
> +        clang-cl reported a compilation error for MSVC bug workaournd code of optionTypeSpecificIndex.

/workaournd/workaround/

> Source/JavaScriptCore/runtime/Options.cpp:294
> +#if COMPILER(CLANG) // for clang-cl

Is both COMPILER(CLANG) and COMPILER(MSVC) true at the same time for MSVC with Clang backend?  Note that this snippet of code is wrapped in a #if COMPILER(MSVC).

[Mark Lam]

Comment on attachment 381277 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=381277&action=review

> Source/JavaScriptCore/runtime/Options.cpp:297
>      size_t index;

I wonder if we can just change this to "size_t index = 0;" and keep the "index = 0;" below as well.  Would that appease both clang and the MSVC backend?

[Fujii Hironori]

Comment on attachment 381277 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=381277&action=review

>> Source/JavaScriptCore/runtime/Options.cpp:294
>> +#if COMPILER(CLANG) // for clang-cl
> 
> Is both COMPILER(CLANG) and COMPILER(MSVC) true at the same time for MSVC with Clang backend?  Note that this snippet of code is wrapped in a #if COMPILER(MSVC).

Yes. clang-cl compiles the almost same code path with MSVC.

>> Source/JavaScriptCore/runtime/Options.cpp:297
>>      size_t index;
> 
> I wonder if we can just change this to "size_t index = 0;" and keep the "index = 0;" below as well.  Would that appease both clang and the MSVC backend?

Will try.

[Fujii Hironori]

Created attachment 381281 [details]
Patch

[Fujii Hironori]

Committed r251276: <https://trac.webkit.org/changeset/251276>

[Radar WebKit Bug Importer]

<rdar://problem/56400312>