g++ 4.8.3 amd64 with MinGW64 toolchain (build fails)

[Mathieu SCHROETER]

Hello,

I try to build wpkg (0.9.4 with patch for MSYS/MinGW support) with MinGW64 and g++ 4.8.3, but :

[ 59%]
Building CXX object libdebpackages/CMakeFiles/debpackages.dir/wpkg_control_fields.cpp.obj
s:/devel/bootstrap-windev/sysroot/toolchain/var/tmp/src/wpkg_0.9.4/libdebpackages/wpkg_control_fields.cpp: In member function 'virtual void wpkg_control::control_file::field_description_t::verify_value() const':
s:/devel/bootstrap-windev/sysroot/toolchain/var/tmp/src/wpkg_0.9.4/libdebpackages/wpkg_control_fields.cpp:884:25: error: call of overloaded 'arg(size_t&)' is ambiguous
                  .arg(len)
                          ^

Regards,

Mathieu

[Mathieu SCHROETER]

In order to fix I cast to (int) because it's just for logging; then it's safe.

[Alexis Wilke]

Maybe we can fix that problem by changing the preprocessor test of the wpkg_output object:

In the header file:

#if defined(MO_WINDOWS) && defined(_WIN64)
    log& arg(const size_t v);
#endif

And then in the cpp file:

#if defined(MO_WINDOWS) && defined(_WIN64)
/** \brief Replace an argument with the size_t value.
 *
 * This function replaces an argument in the format string with the
 * numeric value of the specified parameter.
 *
 * \param[in] v  The value to replace the corresponding argument in the format string.
 *
 * \return The log reference.
 */
log& log::arg(const size_t v)
{
    std::ostringstream ss;
    ss << v;
    f_args.push_back(ss.str());
    return *this;
}
#endif

That should work.

Maybe it will change with a different version of g++ although I use 4.8.1 right now under Ubuntu and I don't have the problem, so I could imagine that it is related to compiling under MinGW and maybe somewhat emulating the behavior of cl.

[Mathieu SCHROETER]

Thank you, see https://sourceforge.net/p/unigw/master/merge-requests/

[Mathieu SCHROETER]

Hello,

Something looks wrong with my patch?

Regards

[Alexis Wilke]

Hi Mathieu,

I'm more that something wrong with my free time...

Also, I think we need to have an MO_MINGW64 instead of using the internal value within the code. And not define MO_MINGW32 when MO_MING64 is defined.

Thank you.
Alexis

[Yevgeny Krasovsky]

Hi guys.
I have created some patches to fix few issues with compilation on MinGW 32 and 64 bit platform.
I tried to create a pull request but did not get it to success.
And thus attached here.

[Yevgeny Krasovsky]

Alexis,

I think there is no need to specify either MO_MINGW32 or MO_MINGW64.
It would make things more complicated with no additional profit.
I believe one define is enough and if you really need to test against 64-bit, then you may check if _WIN64 is defined.

With regards,
Yevgeny

[Mathieu SCHROETER]

Hi,

I use your patches (1, 3 and 5), then it works great with w64 and "MinGW Makefile" generator.

Thank you very much.

[Yevgeny Krasovsky]

Mathieu,
You are welcome.
It would be great if wpkg developers took part in wpkg development too :)

[R. Douglas Barbieri]

And we would love to, as well, but the problem is, we are not getting paid to work on this anymore.