This is done by marking everything that is already loaded but not 'processed'.
After the shim engine is done initializing, the original state is restored,
with the exception of the shim engine itself.
CORE-15846
MSVC was previously given a "result" variable to copy the fscale result from st(0). This led to another "fld" FPU stack push at the very end without popping the source value from the FPU stack.
Moreover, this copy isn't even needed: A simple "fstp st(1)" at the end pops an element from the FPU stack while effectively storing the result in st(0), the register used for returning a double value.
This problem didn't affect GCC, as it is only given the "fscale" instruction and does all necessary stack operations itself.
However, looking into the CRT sources, I found many other i386 implementations with inline assembly suffering from the same problem.
Fortunately, they have been replaced by pure assembly implementations a while ago, so it's time to finally remove them.
ldexp would have also been a candidate for a pure assembly implementation, but the required check for NaN and setting errno (verified on Win2003) already outweighs the benefits.
And we cannot just do a NaN check with FUCOMI as this is an i686/pentiumpro instruction while we're still targeting i586/pentium.
I'm also using this opportunity to clean up the ldexp.c header and only put in the remaining contributors as returned by "git blame".
Thanks to NightWolve1975 for reporting the problem! (https://twitter.com/nightwolve1975/status/1099042477531643912)
Clang-CL does not support #pragma function like CL. The alternative to this approach is to disable all intrinsics for this single source file, but that is not trivial to do, as it can only be done by removing a compiler switch, which we have no CMake support for. Therefore this solution is simpler and as good.
MSBuild, via the VS IDE, uses response files when calling CL or LINK.
We cannot specify a custom response file on the linker command-line,
since specifying response files from within response files is forbidden.
We therefore have to pre-process, at configuration time, the linker
script so as to retrieve the custom linker options to be appended
to the linker command-line.
Running out of pool is likely at least during kmtest:ExPools.
There is a chance of crashing when dereferencing these null pointers -- but
worse, there's also a chance of overwriting the IVT or BDA if a VDM BIOS
call is in progress, which can lead to crashes in non-obvious places later.
Modify the .code16 macro so that the 16-bit code segment can get a
chance to be correctly merged with other (possibly 32-bit) code
segments, without keeping generating multiple .text segments with
different attributes and generating the corresponding LNK4078 warning.
This fixes the warning when compiling NTOSKRNL on MSVC:
"v86.S.obj : warning LNK4078: multiple '.text' sections found with different attributes (C0520040)".
