Fixed warning cast between incompatible function types from FARPROC_ to gettickcount64_t

[kivadiu]

On Windows, GetProcAddress must be cast to something useful which
generates a warning with gcc. This patch silents the warning.

Fixes issue #225

[Lastique]

Pragma diagnostic push/pop appeared only since gcc 4.6. You should check for the compiler version.

Also, you don't need Boost.Predef to detect gcc. You can use BOOST_GCC from Boost.Config. Also, it is quite possible that clang also emits a similar warning, you can check for it with BOOST_CLANG.

[kivadiu]

I cannot check for clang as I do not cross-compile to Windows with clang but I will use BOOST_GCC.

[Lastique]

This warning is only documented (and likely implemented) since gcc 8.

[kivadiu]

That's right, thanks. I will change that.

[kivadiu]

I am very surprised by test failures in completely unrelated code.
For example, it fails with msvc-11 (but not with 9, 10 and 12) at test/threads/thread/members/try_join_for_pass.cpp:147:

boost::thread t0( (G()));
BOOST_TEST(t0.joinable());
t0.try_join_for(boost::chrono::milliseconds(250));
try
{
  t0.join();
  BOOST_TEST(false); // line 147: test failure here
}
catch (boost::system::system_error& e)
{
  BOOST_TEST(e.code().value() == boost::system::errc::invalid_argument);
}

Is it possible that the test server is so heavily loaded that 250 ms are not enough and t0.join() does not fail as expected?

[Lastique]

It is possible, as with any other time-based test.

[viboes]

Andrey is right. This kind of tests can unfortunately fail.

[kivadiu]

How do you normally decide if a patch is harmless then?

[viboes]

I need to check the kind of issue :(