Re: Documentation/MyFirstObjectWalk: add #include "trace.h" to use trace_printf()

From: Junio C Hamano <gitster@pobox.com>
Date: Jun 29, 2023, 19:28 UTC
Message-ID: <xmqq1qhu9ifp.fsf@gitster.g>
In-Reply-To: <CAJoAoZ=OEfsgkqsag926tH4GEuafX26A09SGZ1vR1uLh2W_4TA@mail.gmail.com>

Emily Shaffer <nasamuffin@google.com> writes:

> Yeah, it's almost certainly stale in MyFirstObjectWalk - there was
> very recently a patch to clean up some headers which probably were
> implicitly including trace.h when I wrote this walkthrough.

We are lucky that we have folks like Vinayak who tried out the examples and then bothered to spend time reporting the failure discovered. What does it take, however, for us to have a bit more automated way to prevent such a breakage that comes from API changes? Is it feasible, for example, to add a test that extracts code snippets from the MyFirstObjectWalk document and try to build the result? Alternatively, we can ship such a set of sample source files somewhere in our tree (e.g. contrib/examples?) and have such a test try to build using the current set of source files, but then we need a mechansim to ensure that the sample source files will not go out of sync with the document.

Thoughts?

Thanks.