Lecture 1 — How to use this course¶
Course: Yocto guide · Previous: Course hub · Next: Lecture 02
Build one small robot-observation package, install it into a Linux image, and prove that the image and SDK belong together. Start with a CPU reference so packaging failures do not get confused with algorithm failures.
The project you will carry through the course¶
The supplied observation-check program accepts capture time, completion time, and an age limit in milliseconds. It emits one JSON result and returns a meaningful exit code. It teaches a small part of the observation contract; it is not an inference engine or a complete robot supervisor.
C source → recipe → target package → custom image → boot → target smoke test
↑ ↑
product layer matching SDK + manifest
An executable working on your laptop proves the C behavior. A package build proves a cross-build and packaging step. A booted image with a passing command proves target integration. Keep these as separate evidence levels.
Pick the amount of work your machine supports¶
| Route | Work you can finish | Claim at the end |
|---|---|---|
| Native CPU route | Compile/run the supplied C program, inspect metadata, plan the image and debug fixtures | Native checks and design exercise |
| Metadata route | Also initialize pinned Poky, inspect layers, parse recipes and effective variables | Metadata validation; no target package/image claim |
| Image route | Also build, boot QEMU, generate SDK, run target checks | A demonstrated image and SDK on the named target |
See the lab files and validation status. First-build downloads and toolchain compilation can take hours. Budget setup separately from each 30–60 minute lesson. Stop a build when resources become insufficient or a real failure occurs; save the first failure and resume after fixing it. Do not delete caches as the first response to a problem.
Entry check¶
Explain source script versus running ./script, a compiler versus an executable, and a Git commit versus a moving branch. If these are unfamiliar, complete Lecture 2 before running build commands. Also run the native package checks from the lab README; they should pass before you involve BitBake.
How to record progress¶
Keep source, your layer, configuration, commands, logs, target output, and a revision record in one project. Use separate folders for inputs and generated outputs. Mark each result native tested, metadata tested, target tested, or not run. A screenshot of a successful terminal is useful context, but the actual command and artifact are what another learner can reproduce.
After each session, answer its small self-check without looking. Return two days later and explain the source-to-target path to someone else. Your final assessment is the release-bundle capstone in Lecture 14, using the course rubric.
Answer guidance: source changes the current shell environment; executing a script changes its child process. A cross-compiler runs on the host but produces target code. A branch can move; a recorded full commit identifies one source revision.
Previous: Course hub · Next: Lecture 02