This is the branch of the Headless VM used for Pharo. This branch has the modifications required to run Pharo in a true headless environment. The image running on the VM is responsible of handling the UI and the events.
The image includes a default implementation of the handling of UI through the use of SDL2.
For more details about the whole Pharo VM project refer to our wiki.
This is a fork of OpenSmalltalk-vm. We are doing our best to keep compatibility and contribute back, as long as it fits the objective of Pharo community.
This project is continuously built and test in the CI infrastructure located at:
https://ci.inria.fr/pharo-ci-jenkins2/job/pharo-vm/job/headless/
For building the VM it is required the following set of tools:
- A working Pharo
- CMake (at least version 2.8.4)
- CLang
- Binutils (make and friends)
- wget
- unzip
- automake
- libtool
In Linux Fedora, it is needed to install libcurl and to create a symbolic link to alias such library with the name used by libGit. For doing so, it is required to do:
sudo ln -s /usr/lib64/libcurl.so.4 /usr/lib64/libcurl-gnutls.so.4
In Ubuntu 20.04 and in Mint 20, the VM is built with the following packages:
- build-essentials
- gcc
- g++
- binutils
- cmake
- git
- wget
- unzip
- uuid-dev
- libssl-dev
Building in OSX / Linux:
$ cmake .
$ make install
Building in Windows:
The build in Windows, uses Cygwin. This tool should be installed, and the following Cygwin packages are needed:
- cmake
- mingw64-x86_64-clang
- zip
- unzip
- wget
- curl
- make
- git
$ cmake .
$ make install
The VM is built from generated code and code written by hand. The generated code is the result of converting Smalltalk code into C. This conversion is performed during the cmake process.
This will generate the VM in build/dist/
By default the cmake build will build a CogVM, that is, a VM configuration with JIT compilation. Our cmake configuration accepts a FLAVOUR
argument to specify different vm flavours to build, which you can use as follows:
$ cmake -DFLAVOUR=[your flavour] .
The accepted flavours for the moment are as follows:
- CoInterpreterWithQueueFFI: VM including JIT
- StackVM: VM with context to native stack mapping, without JI
The headless mode is developed on top of code of the Cog branch of Opensmalltalk-vm. The code that is used without changes is stored in the extracted directory. This allows us to easy integrate changes from and to the Cog branch.
The code that has been specially created or modified for this branch is stored in src / include and plugins.
- smalltalksrc: includes the tonel repository with the code in Slang.
- generated: here VMMaker will generate the VM code.
- includes: All non generated includes required by headless mode
- src: All non generated code required by headless mode.
- extracted: This code is literally the same code base used in the normal OpenSmalltalk-VM build.
- plugins: The code of the different plugins developed for headless mode.
You can load the Pharo code of the VM using Pharo's git client Iceberg. You can do so by cloning directly this repository from Iceberg, or by adding an already existing clone to it.
Alternatively, if you're building the VM using the instructions above, the build process does already generate a Pharo image with the VM code loaded. You'll find such Image in the build/vmmaker
directory inside your build directory.