clang doesn't make the difference between UL & ULL sufixes

[bog-dan-ro]

thread/qthread.h:108:36: error: implicit conversion from 'unsigned long long' to 'unsigned long' changes value from 18446744073709551615 to
4294967295 [-Werror,-Wconstant-conversion]
bool wait(unsigned long time = ULONG_MAX);
~ ^~~~~~~~~
/home/bogdan/necessitas/android-ndk/platforms/android-9/arch-arm/usr/include/sys/limits.h:72:20: note: expanded from macro 'ULONG_MAX'

define ULONG_MAX 0xffffffffffffffffUL

                    ^~~~~~~~~~~~~~~~~~~~

Full command line:
/home/bogdan/necessitas/android-ndk/toolchains/llvm/prebuilt/linux-x86_64/bin/clang++ -c -target armv7-none-linux-androideabi -march=armv7-a -mfpu=vfp -mfloat-abi=softfp -ffunction-sections -funwind-tables -fstack-protector -fno-short-enums -DANDROID -Wa,--noexecstack -fno-builtin-memmove -D__LP64__ --sysroot=/home/bogdan/necessitas/android-ndk/platforms/android-9/arch-arm/ -g -Os -fomit-frame-pointer -fno-strict-aliasing -mthumb -std=c++1z -fvisibility=hidden -fvisibility-inlines-hidden -Wall -W -Wdate-time -D_REENTRANT -fPIC -DQT_NO_MTDEV -DQT_NO_LIBUDEV -DQT_NO_TSLIB -DQT_NO_LIBINPUT -DQT_NO_USING_NAMESPACE -DQT_HAVE_POLL -DQT_BUILD_CORE_LIB -DQT_BUILDING_QT -DQT_NO_CAST_TO_ASCII -DQT_ASCII_CAST_WARNINGS -DQT_MOC_COMPAT -DQT_USE_QSTRINGBUILDER -DQT_DEPRECATED_WARNINGS -DQT_DISABLE_DEPRECATED_BEFORE=0x050000 -DQT_NO_DEBUG -Wall -Wextra -Werror -Woverloaded-virtual -Wshadow -Wundef -Wnon-virtual-dtor -Wpointer-arith -Wformat-security -Wno-long-long -Wno-variadic-macros -pedantic-errors -Wchar-subscripts -Wold-style-cast -I. -I/home/bogdan/work/qt/openssl-1.0.2f/include -Iglobal -I../3rdparty/pcre -I../3rdparty/harfbuzz/src -I../3rdparty/md5 -I../3rdparty/md4 -I../3rdparty/sha3 -I../3rdparty/double-conversion/include -I../3rdparty/double-conversion/include/double-conversion -I../3rdparty/forkfd -I../../include -I../../include/QtCore -I../../include/QtCore/5.7.0 -I../../include/QtCore/5.7.0/QtCore -I.moc -isystem /home/bogdan/necessitas/android-ndk/sources/cxx-stl/gnu-libstdc++/4.9/include -isystem /home/bogdan/necessitas/android-ndk/sources/cxx-stl/gnu-libstdc++/4.9/libs/armeabi-v7a/include -isystem /home/bogdan/necessitas/android-ndk/platforms/android-9/arch-arm/usr/include -I../../mkspecs/android-clang -DQT_NO_CAST_TO_ASCII=1 -DQT_NO_CAST_FROM_ASCII=1 -DQT_STRICT_ITERATORS -DQT_NO_URL_CAST_FROM_STRING=1 -DQT_NO_CAST_FROM_BYTEARRAY=1 -DQT_NO_KEYWORDS=1 -DQT_USE_QSTRINGBUILDER -DQT_USE_FAST_OPERATOR_PLUS -Dsignals=int -Dslots=int -Demit=public: -Dforeach=public: -Dforever=public: -xc++ thread/qthread.h -o .obj/header_qthread.o
thread/qthread.h:108:36: error: implicit conversion from 'unsigned long long' to 'unsigned long' changes value from 18446744073709551615 to
4294967295 [-Werror,-Wconstant-conversion]
bool wait(unsigned long time = ULONG_MAX);
~ ^~~~~~~~~
/home/bogdan/necessitas/android-ndk/platforms/android-9/arch-arm/usr/include/sys/limits.h:72:20: note: expanded from macro 'ULONG_MAX'

define ULONG_MAX 0xffffffffffffffffUL

                    ^~~~~~~~~~~~~~~~~~~~

1 error generated.

[bog-dan-ro]

Ah, is -D__LP64__ fault, sorry for the noise!

[bog-dan-ro]

Actually I still think it is a problem, xxxxUL should be unsigned long no matter what

[DanAlbert]

You'd think so, but the standard has other ideas: http://en.cppreference.com/w/cpp/language/integer_literal#The_type_of_the_literal

The L suffix means "at least long", but if the literal is large enough it can still promote to long long.

[DanAlbert]

What is really weird is that you're getting the LP64 definition of ULONG_MAX for armeabi-v7a. I can't manage to make that happen with my test case. That definition is guarded by #ifdef __LP64__.

[bog-dan-ro]

I added manually -D__LP64__ to command line, because with clang failed to cast

GL_APICALL void GL_APIENTRY glCompressedTexSubImage2D (
    GLenum target, GLint level, GLint xoffset, GLint yoffset,
    GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const GLvoid* data);

to

void (*CompressedTexSubImage2D)(
    GLenum target, GLint level, GLint xoffset, GLint yoffset,
    GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const GLvoid *data)

opengl/qopengltexturehelper.cpp:160:29: error: assigning to 'void (*)(GLenum, GLint, GLint, GLint, GLsizei, GLsizei, GLenum, GLsizei, const GLvoid *)' (aka 'void (*)(unsigned int, int, int, int, int, int, unsigned int, int, const void *)') from incompatible type 'void (GLenum, GLint, GLint, GLint, GLsizei, GLsizei, GLenum, GLsizei, const GLvoid *) __attribute__((pcs("aapcs")))' (aka 'void (unsigned int, int, int, int, int, int, unsigned int, int, const void *) __attribute__((pcs("aapcs")))')
CompressedTexSubImage2D = ::glCompressedTexSubImage2D;
                            ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~

[DanAlbert]

If you manually add -D__LP64__ on a 32-bit system, you're gonna have a bad time. A really, really bad time.

Looks like the incompatibility in the case is just the __attribute__((aapcs))? Are you using armeabi-v7a-hard?

I think the problem here is that you need:

-void (*CompressedTexSubImage2D)(
+GL_APICALL void (*CompressedTexSubImage2D)(
    GLenum target, GLint level, GLint xoffset, GLint yoffset,
    GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const GLvoid *data);

[bog-dan-ro]

If you manually add -D__LP64__ on a 32-bit system, you're gonna have a bad time. A really, really bad time.

Yup, I've seen it :)

Looks like the incompatibility in the case is just the attribute((aapcs))? Are you using armeabi-v7a-hard?

Yep "attribute((aapcs))" is the problem, I just don't know why gcc didn't warned me about it... or this is a false positive clang error, considering that function signature has no floating-point params ...

No, I'm not using v7a-hard, I'm using "-target armv7-none-linux-androideabi -march=armv7-a -mfpu=vfp -mfloat-abi=softfp ....".

[bog-dan-ro]

@DanAlbert just to be crystal clear: __NDK_FPABI__ should be defined as __attribute__((pcs("aapcs"))) when compiles with "-target armv7-none-linux-androideabi -march=armv7-a -mfpu=vfp -mfloat-abi=softfp" flags?

[DanAlbert]

__NDK_FPABI__ can be empty for -mfloat-abi=softfp. That's the default.

The hard float ABI is going away with r12. You should be able to ignore __NDK_FPABI__.

[bog-dan-ro]

Can be or should be? :)
If should be, then you should reopen this issue because, I just test it with r12 beta2 and clang still defines __attribute__((pcs("aapcs"))) ;-(

[DanAlbert]

I guess I forgot to actually hit the re-open button, but I did actually make another pass at this in r13. All the aapcs stuff has been completely purged from our headers now, so everything should Just Work.