What is wrong
src/CanKit.Pro.IsoTp/README.md, section Timing accuracy — STmin pacing (NFR-003), ends with:
Verified end-to-end by tests/CanKit.Pro.Tests/TestCases/IsoTp/IsoTpStminTimingTests.cs (Virtual-loopback CF-spacing measurement with CI-tolerant soft bounds).
IsoTpStminTimingTests no longer does that. Its class documentation says the pacing assertion "used to be a soft wall-clock bound" and "now runs on a clock the test drives, so the property asserted is the one the code implements and there is no tolerance to widen", and that "nothing in this class gates on" effective spacing on a real runner (see #92). The test advances an injected clock and brackets each interval exactly.
So the README claims a real-time spacing measurement with tolerance bounds, and the test that the sentence points at makes no real-time measurement at all.
The sentence before it has the same shape: "typically within ±1 ms of the configured value on idle Windows/Linux/macOS hosts". Nothing in this repository measures that either; if it comes from an observation, the README should say where.
Found by
Codex review on #245, which added a What is validated, and what is not section to every package README and had first repeated this paragraph's wording in the IsoTp one. That section is corrected in that PR; this paragraph predates it and was left out of it, because it is not something that PR introduced.
Suggested change
Say what the test verifies (the sender releases one Consecutive Frame per STmin on a clock the test drives), say that effective real-time spacing is STmin plus scheduling latency and is not measured by the suite, and either source the ±1 ms figure or drop it.
Documentation only.
What is wrong
src/CanKit.Pro.IsoTp/README.md, section Timing accuracy — STmin pacing (NFR-003), ends with:IsoTpStminTimingTestsno longer does that. Its class documentation says the pacing assertion "used to be a soft wall-clock bound" and "now runs on a clock the test drives, so the property asserted is the one the code implements and there is no tolerance to widen", and that "nothing in this class gates on" effective spacing on a real runner (see #92). The test advances an injected clock and brackets each interval exactly.So the README claims a real-time spacing measurement with tolerance bounds, and the test that the sentence points at makes no real-time measurement at all.
The sentence before it has the same shape: "typically within ±1 ms of the configured value on idle Windows/Linux/macOS hosts". Nothing in this repository measures that either; if it comes from an observation, the README should say where.
Found by
Codex review on #245, which added a What is validated, and what is not section to every package README and had first repeated this paragraph's wording in the IsoTp one. That section is corrected in that PR; this paragraph predates it and was left out of it, because it is not something that PR introduced.
Suggested change
Say what the test verifies (the sender releases one Consecutive Frame per STmin on a clock the test drives), say that effective real-time spacing is STmin plus scheduling latency and is not measured by the suite, and either source the ±1 ms figure or drop it.
Documentation only.