Mitsubishi Electric Research Laboratories

DiamondBuild

DiamondBuild is a portable build system designed to enable professional and reliable software management in a highly heterogeneous and diverse software development environment such as the one present at MERL.  DiamondBuild provides a lightweight framework that enables effective software sharing, integration, and delivery to our customers.
     Although the system has been driven by MERL's own complex software management problems, these problems are common in software development, especially in the open-source community.  The DiamondBuild project has synergy with existing open-source efforts to develop a portable and open build system.  The ultimate objective of this project is to provide a mature prototype and specification to bolster this open-source effort, which MERL can then benefit from.

Background & Objective:  MERL's software development environment is characterized by usage of multiple operating systems, programming languages, compilers, and third-party open-source software packages.  Furthermore many different versions of these tools are in use concurrently and individual projects follow diverse software development practices, making it difficult to share and integrate software internally and even more difficult to reliably transfer complex projects to our customers.
     The objective of the DiamondBuild project is to establish a portable modular framework and build system that results in good and consistent software development and management practices througout MERL and results in reliable and effective transfer to our customers in Japan.

Technical Discussion:  DiamondBuild has already evolved through a number of manifestations, including an initial version built entirely in "Make" and several versions augmented with "Perl".  DiamondBuild has enabled software porting, sharing, management, and delivery of a number of software projects.  Various usage problems associated with performance (re-build speed), reliability (due to OS-related differences), and absolute correctness (various unusual usage cases) remain to be solved before true widespread use is promoted.

Technology Area:  Computer Vision

Modification Date:  January 22, 2007