Restructure the compilation process


The installation process looks very Byzantine to people who are not so deep into OpenCPI themselves – that is, anyone who is not an OpenCPI developer. We would be very very very interested in using OpenCPI in several projects, and also to contribute to it, but so far no one in our team (4 experienced embedded engineers) has managed to have an installation working (and we already invested more than 1w/m).
This is due to the huge amount of scripts involved in the build process, and the absence of up-to-date guidelines.
I am pretty sure that redoing the build system from scratch (e.g., by using CMake instead of the whole mess of makefile scripts) is far from being a trivial thing, but I am sure as well that this is exactly the reason why OpenCPI does not get all the attention it deserves — it is pretty difficult to get excited about something that takes 2 weeks and countless abuses to be installed...
It would be already extremely helpful to have a single configuration file, in text format (JSON, XML, ...) and well commented, for the installation, where all the variables and options are clearly stated.
This would also make testing much easier from your side — simply set up a machine running 24/7 that fills this file with random configuration combinations, and take note of those failing...




OpenCPI Guest
May 3, 2018, 5:21 AM

Normally no need to perform the compilation by hand, RPMs are available on !!

James Kulp
May 15, 2018, 9:25 PM

The source build system is being reworked at this moment, but there are RPM-based installations that simply require a few yum commands. See

James Kulp
May 15, 2018, 9:26 PM

The new source build system is being tested with single-command installations across a range of OS configurations. Stay tuned.




OpenCPI Guest