Dear colleagues,
We are researchers at Inria (France) working on software configurable systems in general.
Recently we have made lots of effort with Linux: we have basically build 95K+ configurations using ranconfig.
Our attempt is quite similar to KernelCI, the platform you’ve developed.
We have developed some techniques to identify what (combinations of) options lead to a failure (a non-trivial problem!), with two main applications in mind: (1) configuration bug location/understanding (2) preventing configurations’ build that will anyway lead to a failure. We mainly use randconfig.
More details can be found here:
We have also made some efforts about predicting kernel sizes:
It would be very nice to know more about your infrastructure, your strategy for sampling configurations, and basically your insights. We can also chat a bit about our approach and potential complementarity.
Would you be available for discussing?
I guess some of you will be at OSS in Lyon.
I’m giving a talk on Wednesday
Best regards,