Working With Multiple Sandboxes

You might want to develop code in more than one sandbox. This is particularly useful for testing in a separate sandbox from your development sandbox.

Note that you do not need multiple sandboxes to work on multiple versions of a project. Program development directories (under project/ and common/) can be located anywhere and there can have replicated variations. In this case the variants all share the same install/ and import/ directories.

You can create a new sandbox in an arbitrary directory with by running the following command. Note that this has to be run from a Linux system, but will establish a sandbox for Windows as well.

> /research/graphics/tools/linux/install_linux/bin/setup/new.g.user.sh  <Sandbox/>

The Sandbox/ directory will now contain:

install_linux/        Installed copies of Linux software developed at Brown.
install_WIN32/        Installed copies of Windows software developed at Brown.
common/               The most recent common code checked out from CVS.
project/              An empty directory for your project code.

Third party software for all platforms will be accessed for all sandboxes from: /research/graphics/tools/linux/import/

In order to switch between sandboxes you can designate the active one by setting the G environment variable to the active sandbox. This variable is used to access the install_linux/ directory and the make files in common/build/make. Rather than switching the G variable it is probably more convenient to open two terminal windows each with different settings for G.

The GTO variable determines where software developed at Brown is installed and accessed. If you want to have totally independent sandboxes then you need to set it to the value of G for the active sandbox. You can also have multiple install directories for a single sandbox. In this case GTO can be changed to point to one of several target install directories.