Skip to content
This repository has been archived by the owner on Aug 18, 2022. It is now read-only.

Latest commit

 

History

History
120 lines (95 loc) · 4.5 KB

BUILD_WIN.md

File metadata and controls

120 lines (95 loc) · 4.5 KB
Raw

Build Instructions for Windows

There are two main options for compiling matlab-json on Windows:

  1. Build json-c from source
  2. Use pre-built json-c library

Instructions for both options follow.

Build from Source

Prerequisites

In order to compile json-c and matlab-json, you will need the following tools:

Code

The make file for Windows assumes that the json-c source code is available in a directory called json-c within the same parent directory as the directory containing the matlab-json source code.

A typical layout would be:

  • src/
    • json-c/
      • json.h
      • ...
    • matlab-json/
      • make.m
      • fromjson.c
      • ...

Compile JSON-C

  • Open json-c.vcxproj file in Visual Studio 2013
  • Select "Debug" configuration
  • Select appropriate platform for the target MEX file (win32 => mexw32, x64 => mexw64)
  • In the project properties, set the following properties
    • "Configuration Properties\General\Platform Toolset" = "Windows7.1SDK"
  • Check the following project properties:
    • "Configuration Properties\General\Output Directory" = "$(SolutionDir)$(Configuration)" (win32 only)
    • "Configuration Properties\General\Output Directory" = "$(SolutionDir)$(Platform)$(Configuration)" (x64 only)
    • "Configuration Properties\General\Target Name" = "$(ProjectName)"
    • "Configuration Properties\General\Target Extension" = ".lib"
    • "Configuration Properties\General\Configuration Type" = "Static Library"
    • "Configuration Properties\C\C++\Advanced\CompileAs" = "Compile as C++ Code (/TP)"
    • "Configuration Properties\C\C++\Code Generation\Runtime Library" = "Multi-threaded DLL (/MD)"
    • "Configuration Properties\Librarian\General\Output File" = "$(OutDir)$(TargetName)$(TargetExt)"
  • Build the project

The json-c library file json-c.lib should have been created in the location that is expected by the matlab-json make function.

Compile MATLAB-JSON

  • Open MATLAB and change to the matlab-json directory
  • Run mex -setup, if this hasn't been done before and you're using R2013b or earlier, and select "Microsoft Software Development Kit (SDK) 7.1"
  • Run make

Use Pre-built Library

This approach relies on the Appveyor build scripts added to the json-c project in Pull Request #200 to build the json-c library in a Continuous Integration (CI) environment, rather than installing the necessary tools and compiling it locally.

As of 5-Nov-2015, this pull request has not yet been merged and there is no official or semi-official CI project for json-c, but the CI build script in this pull request can still be used.

Prerequisities

The only prerequisite tools for this option is the Windows SDK 7.1 (see above).

Obtain pre-built library

Since there is no official or semi-official CI project for json-c, there are two main options for obtaining the pre-built library for this approach:

  1. Create your own fork of json-c that merges the required pull request and any others that are required to make the Windows version build correctly, and create your own Appveyor project.
  2. Use someone else's version of option 1 such as this one. In this version, as of 5-Nov-2015, the ci_dev branch is the most up-to-date.

From the Appveyor project:

  • Navigate to the appropriate (passing) build
  • Open the job for "Debug" configuration for "Windows7.1SDK"
  • Open the "Artifacts" tab
  • Download the artifact json-c.lib.zip

Extract pre-built library

The contents of the json-c.lib.zip file described above should be extracted into the root directory of the matlab-json source. The resulting directory structure will be as follows:

  • matlab-json/
    • include/
      • json-c/
        • json.h
        • ...
    • lib/
      • json-c.lib
    • make.m
    • fromjson.c
    • ...

Compile MATLAB-JSON

The steps for compiling matlab-json itself are the same as the other approach (see above).

The make script will detect the compiled library and header files in the matlab-json directory and use these in preference to any that are in the json-c source directory (if it exists).