Hi Greg and Anthony, On 3/21/21, Gregory Nutt <spudan...@gmail.com> wrote: > >> ... what we can >> do to prevent these types of "side-effects" when people are touching the >> config system. > > The use of defconfig files implies that we never change default values > of configuration. If someone does change the default value of a > configuration setting, then it effects ALL configurations that depend on > that default. We need to catch this is the review step. During > review, if we notice that there is a change to the default value of the > configuration setting, then we must insist that all defconfig files be > modified so that the previous configuration is not changed due to the > side-effect. > > CI cannot catch this. As noted, the error does not occur until run > time. A proper regression test would catch this but we do not do any > automated regression testing. That is a major hole in the testing. >
Yes, the CI cannot catch it, but using some QA Unit Tests running in the Linux simulator or QEMU we could catch it. Abdelatif tested the smartfs example in the "sim" board and the issue also happened. Of course it cannot catch all the issues, but at least most of the issues could be avoided using some basic tests. BR, Alan
