view doc/README.Windows @ 457:90b1d927912a

- suite_aggrs: make sure random memory used doesn't ever result in NaN fp values (for every possible address), as this messes with result comparison
author Tassilo Philipp
date Fri, 28 Jan 2022 14:11:21 +0100
parents 451299d50c1a
children
line wrap: on
line source

Bulding DynCall on Windows
==========================


Build using buildsys/nmake and MS Visual Studio Tools:
------------------------------------------------------

  32-bit:
    vcvarsall.bat x86
    .\configure.bat /target-x86
    nmake /f Nmakefile

  64-bit:
    vcvarsall.bat amd64
    .\configure.bat /target-x64
    nmake /f Nmakefile

  Build/run tests:
    cd test
    nmake /f Nmakefile
    
    NOTE: if you encounter 'python'/'lua' not found messages (which is 
     executed mistakenly for re-generation of test C code)
     timestamps are wrong; workaround: run "svn revert -R ." several times.


Build using buildsys/gmake and MinGW or Cygwin:
-----------------------------------------------

  32-bit:
    .\configure.bat /target-x86 /tool-gcc
    make

  64-bit:
    .\configure.bat /target-x64 /tool-gcc
    make


Build using Makefile.embedded and MinGW or Cygwin:
--------------------------------------------------

    > mingw32-make -f Makefile.embedded
    $ make -f Makefile.embedded CC=gcc


Build using CMake and Nmake:
----------------------------

  Put the appropriate tool-chain in path (32 or 64-bit version of MSVC), then:
  
    cmake -G "NMake Makefiles" %DYNCALL_SOURCE_DIR%
    nmake


Build using CMake and Visual Studio:
------------------------------------
 
  32-bit:
    cmake .
    open DynCall.sln

  BROKEN: SAFE EH.

  BROKEN: Rules to assemble MASM files are still missing.


Build ARM64 using CMake and Visual Studio 2019:
-----------------------------------------------

   vcvarsall.bat amd64_arm64
   mkdir build && cd build
   cmake -G "Visual Studio 16 2019" -A ARM64 ..
   cmake --build . --config Release