On Tue, May 19, 2020 at 1:06 PM Gregory Nutt <spudan...@gmail.com> wrote: > On Tue, May 19, 2020 at 12:57 PM Alan Carvalho de Assis <acas...@gmail.com> > wrote: > > During the NuttX Workshop a friend (Thiago Costa Paiva) suggested > > about the idea of creating a FPGA solution to validate NuttX hardware, > > but his idea didn't take place. > > Adding hardware raises complexities or design, manufacturing, > distribution, and support. Not insumountble, but not simple either.
This is why I suggested that we "grow" into it. (1) Perfect the current coding standard and build test. (2) When #1 is complete, add static analysis. (3) When #2 is complete, add software-only automated test suite under simulation. (4) When #3 is complete, add hardware testing. The recommendation here is to design #3 so that a minimal porting effort turns it into #3 and #4, and makes #4 possible on a variety of hardware. In general: The more toolchains, static analyzers, build hosts, target device types, and boards that are used to exercise the software, the better and more reliable the software will become. > A minor correction: We do not validate hardware, we validate ONLY > software. Any hardware is part of the trusted test harness and not a > part of the unit under test. Software is the unit under test. Correct. Nathan
