Bug 216280 – REGRESSION(r266681) [GTK] constexpr build failure in LTS/Debian with GCC 8

• Component: WebCore Misc.
• Status: RESOLVED FIXED

[Lauro Moura]

Error:

[257/502] Building CXX object Source/WebCore/CMakeFiles/WebCore.dir/__/__/DerivedSources/WebCore/unified-sources/UnifiedSource-3c72abbe-54.cpp.o
FAILED: Source/WebCore/CMakeFiles/WebCore.dir/__/__/DerivedSources/WebCore/unified-sources/UnifiedSource-3c72abbe-54.cpp.o 
<snip long include line>
In file included from DerivedSources/ForwardingHeaders/wtf/StdLibExtras.h:32:0,
                 from DerivedSources/ForwardingHeaders/wtf/FastMalloc.h:26,
                 from ../../Source/WebCore/config.h:56,
                 from ../../Source/WebCore/platform/text/TextCodecCJK.cpp:26,
                 from DerivedSources/WebCore/unified-sources/UnifiedSource-3c72abbe-54.cpp:1:
../../Source/WebCore/platform/text/TextCodecSingleByte.cpp: In function ‘constexpr const SingleByteDecodeTable& WebCore::tableForDecoding(WebCore::TextCodecSingleByte::Encoding)’:
DerivedSources/ForwardingHeaders/wtf/Assertions.h:587:56: error: call to non-constexpr function ‘void CRASH_WITH_INFO(...)’
 #define RELEASE_ASSERT_NOT_REACHED(...) CRASH_WITH_INFO(__VA_ARGS__)
                                         ~~~~~~~~~~~~~~~^~~~~~~~~~~~~
../../Source/WebCore/platform/text/TextCodecSingleByte.cpp:237:5: note: in expansion of macro ‘RELEASE_ASSERT_NOT_REACHED’
     RELEASE_ASSERT_NOT_REACHED();
     ^~~~~~~~~~~~~~~~~~~~~~~~~~

GCC 8 has some limitations checking non-constexpr expressions after return statements.

Example: https://godbolt.org/z/dhPvaf

Related GCC bugs: 
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86678
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67371

[Lauro Moura]

Created attachment 408256 [details]
Patch

[Darin Adler]

Comment on attachment 408256 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=408256&action=review

> Source/WebCore/platform/text/TextCodecSingleByte.cpp:248
> +    default:
> +        RELEASE_ASSERT_NOT_REACHED();

I guess we can do this. But sadly this has a side effect of suppressing the compile-time warning if we forget to include one of the enumeration values here. Would be nice to find a solution that does not do that.

[Alex Christensen]

Comment on attachment 408256 [details]
Patch

I'd prefer to just remove the constexpr from this function.

[Alex Christensen]

Comment on attachment 408256 [details]
Patch

I'd prefer to just remove the constexpr from this function.

[Darin Adler]

Don’t we need the constexpr to make this work as a template argument?

[Alex Christensen]

Comment on attachment 408256 [details]
Patch

No, tableForDecoding isn't used as a template argument.

[Alex Christensen]

Created attachment 408259 [details]
Patch

[Alex Christensen]

Created attachment 408260 [details]
Patch

[Alex Christensen]

Lauro, does this fix your build?

[Lauro Moura]

(In reply to Alex Christensen from comment #9)
> Lauro, does this fix your build?

Could not test it direcly on the bot but indeed it should fix.

[Alex Christensen]

http://trac.webkit.org/r266749

[Radar WebKit Bug Importer]

<rdar://problem/68529158>