aquiles2k
commented
5 months ago So, annulen over at qtwebkit repo confirmed that 32bit JIT was abandoned long time ago which is why this code is falling back to CLoop which again is untested and unsupported on 32bit mingw. So i guess i have to give up on this.
Meanwhile i tried for 64bit mingw crosscompile which got me further but still errors out early.
There were some fixable errors like mispelled or not existing includes.
The first error that left me scratching my head was:
[ 16%] [32mBuilding CXX object Source/JavaScriptCore/CMakeFiles/LLIntOffsetsExtractor.dir/llint/LLIntOffsetsExtractor.cpp.obj[0m
cd /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07.build_/Source/JavaScriptCore && /home/adrian/dev/sources/mxe/usr/bin/x86_64-w64-mingw32.shared-g++ -DBUILDING_LLIntOffsetsExtractor -DBUILDING_QT__=1 -DBUILDING_WEBKIT=1 -DBUILDING_WITH_CMAKE=1 -DHAVE_CONFIG_H=1 -DNOMINMAX -DPAS_BMALLOC=1 -DQT_ASCII_CAST_WARNINGS -DQT_CORE_LIB -DQT_DEPRECATED_WARNINGS -DQT_DISABLE_DEPRECATED_BEFORE=0x050000 -DQT_NO_CAST_TO_ASCII -DQT_NO_DEBUG -DQT_NO_DYNAMIC_CAST -DQT_NO_EXCEPTIONS -DQT_NO_NARROWING_CONVERSIONS_IN_CONNECT -DQT_USE_QSTRINGBUILDER -DSTATICALLY_LINKED_WITH_WTF -DUNICODE -D_UNICODE -D_WINDOWS -D_WINSOCKAPI_="" -D__USE_MINGW_ANSI_STDIO=1 @CMakeFiles/LLIntOffsetsExtractor.dir/includes_CXX.rsp -Wextra -Wall -pipe -fmax-errors=20 -Wno-odr -Wno-stringop-overread -Wno-stringop-overflow -Wno-nonnull -Wno-array-bounds -Wno-expansion-to-defined -Wno-noexcept-type -Wno-psabi -Wno-misleading-indentation -Wno-maybe-uninitialized -Wwrite-strings -Wundef -Wpointer-arith -Wmissing-format-attribute -Wformat-security -Wcast-align -Wno-unknown-pragmas -Wno-tautological-compare -fpermissive -fno-strict-aliasing -fno-exceptions -fno-rtti -mno-ms-bitfields -fno-keep-inline-dllexport -O3 -DNDEBUG -std=c++20 -MD -MT Source/JavaScriptCore/CMakeFiles/LLIntOffsetsExtractor.dir/llint/LLIntOffsetsExtractor.cpp.obj -MF CMakeFiles/LLIntOffsetsExtractor.dir/llint/LLIntOffsetsExtractor.cpp.obj.d -o CMakeFiles/LLIntOffsetsExtractor.dir/llint/LLIntOffsetsExtractor.cpp.obj -c /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/llint/LLIntOffsetsExtractor.cpp
In file included from /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/heap/Heap.h:40,
from /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/runtime/JSCell.h:29,
from /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/runtime/JSCast.h:28,
from /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/runtime/ClassInfo.h:26,
from /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/runtime/Structure.h:28,
from /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/bytecode/ArrayProfile.h:29,
from /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/llint/LLIntOffsetsExtractor.cpp:29:
/home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/heap/IsoSubspace.h:81:23: error: function 'JSC::GCClient::IsoSubspace::~IsoSubspace()' definition is marked dllimport
81 | JS_EXPORT_PRIVATE ~IsoSubspace() = default;
| ^
In file included from /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/debugger/Debugger.h:26,
from /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/runtime/VMInlines.h:28,
from /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/runtime/CommonSlowPaths.h:39,
from /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/llint/LLIntOffsetsExtractor.cpp:33:
/home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/debugger/DebuggerCallFrame.h:57:43: error: function 'const WTF::TextPosition& JSC::DebuggerCallFrame::position() const' definition is marked dllimport
57 | JS_EXPORT_PRIVATE const TextPosition& position() const { return m_position; }
| ^~~~~~~~
make[4]: *** [Source/JavaScriptCore/CMakeFiles/LLIntOffsetsExtractor.dir/build.make:169: Source/JavaScriptCore/CMakeFiles/LLIntOffsetsExtractor.dir/llint/LLIntOffsetsExtractor.cpp.obj] Error 1Since i have read somewhere that dllimport/export markings are more of a msvc thing and not gcc (which is used here), i disabled them:
diff --git a/Source/WTF/wtf/PlatformUse.h b/Source/WTF/wtf/PlatformUse.h
index 1111111..2222222 100644
--- a/Source/WTF/wtf/PlatformUse.h
+++ b/Source/WTF/wtf/PlatformUse.h
@@ -44,7 +44,7 @@
/* Export macro support. Detects the attributes available for shared library symbol export
decorations. */
-#if OS(WINDOWS) || (COMPILER_HAS_CLANG_DECLSPEC(dllimport) && COMPILER_HAS_CLANG_DECLSPEC(dllexport))
+#if (COMPILER_HAS_CLANG_DECLSPEC(dllimport) && COMPILER_HAS_CLANG_DECLSPEC(dllexport))
#define USE_DECLSPEC_ATTRIBUTE 1
#elif defined(__GNUC__)
#define USE_VISIBILITY_ATTRIBUTE 1Can anyone confirm this is the right approach?
Anyway, this got me some compilations further but now im getting an assembler error again:
[ 19%] [32mBuilding CXX object Source/JavaScriptCore/CMakeFiles/JavaScriptCore.dir/assembler/MacroAssemblerX86Common.cpp.obj[0m
cd /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07.build_/Source/JavaScriptCore && /home/adrian/dev/sources/mxe/usr/bin/x86_64-w64-mingw32.shared-g++ -DBUILDING_JavaScriptCore -DBUILDING_QT__=1 -DBUILDING_WEBKIT=1 -DBUILDING_WITH_CMAKE=1 -DHAVE_CONFIG_H=1 -DNOMINMAX -DPAS_BMALLOC=1 -DQT_ASCII_CAST_WARNINGS -DQT_CORE_LIB -DQT_DEPRECATED_WARNINGS -DQT_DISABLE_DEPRECATED_BEFORE=0x050000 -DQT_NO_CAST_TO_ASCII -DQT_NO_DEBUG -DQT_NO_DYNAMIC_CAST -DQT_NO_EXCEPTIONS -DQT_NO_KEYWORDS -DQT_NO_NARROWING_CONVERSIONS_IN_CONNECT -DQT_USE_QSTRINGBUILDER -DSTATICALLY_LINKED_WITH_WTF -DUNICODE -D_UNICODE -D_WINDOWS -D_WINSOCKAPI_="" -D__USE_MINGW_ANSI_STDIO=1 @CMakeFiles/JavaScriptCore.dir/includes_CXX.rsp -Wextra -Wall -pipe -fmax-errors=20 -Wno-odr -Wno-stringop-overread -Wno-stringop-overflow -Wno-nonnull -Wno-array-bounds -Wno-expansion-to-defined -Wno-noexcept-type -Wno-psabi -Wno-misleading-indentation -Wno-maybe-uninitialized -Wwrite-strings -Wundef -Wpointer-arith -Wmissing-format-attribute -Wformat-security -Wcast-align -Wno-unknown-pragmas -Wno-tautological-compare -fpermissive -fno-strict-aliasing -fno-exceptions -fno-rtti -mno-ms-bitfields -fno-keep-inline-dllexport -O3 -DNDEBUG -std=c++20 -ffp-contract=off -MD -MT Source/JavaScriptCore/CMakeFiles/JavaScriptCore.dir/assembler/MacroAssemblerX86Common.cpp.obj -MF CMakeFiles/JavaScriptCore.dir/assembler/MacroAssemblerX86Common.cpp.obj.d -o CMakeFiles/JavaScriptCore.dir/assembler/MacroAssemblerX86Common.cpp.obj -c /home/adrian/dev/sources/mxe/tmp-qtwebkit2-x86_64-w64-mingw32.shared/webkit-2022-09-07/Source/JavaScriptCore/assembler/MacroAssemblerX86Common.cpp
{standard input}: Assembler messages:
{standard input}:131: Error: unknown pseudo-op: `.previous'
make[4]: *** [Source/JavaScriptCore/CMakeFiles/JavaScriptCore.dir/build.make:964: Source/JavaScriptCore/CMakeFiles/JavaScriptCore.dir/assembler/MacroAssemblerX86Common.cpp.obj] Error 1
make[4]: *** Waiting for unfinished jobs....Here we are again with some compiler voodoo i dont understand. I suspect there will be some defines or compiler flags which would tell the assembler to make the right things, but im lost here for now.
Anyone else an idea?
quarcko
Im trying to build the latest qt5 code with mxe crosscompiled for win.
After some initial patches it seems its getting configured but fails later with the following:
Has anyone any clue whats going wrong? Is it perhaps some misconfiguration? Some missing includes? This bytecode generation voodoo is so low level - i dont have a clue where to look for fixing this.