On Mon, Oct 28, 2019 at 10:24:31AM +0900, Junio C Hamano wrote:
> SZEDER Gábor <szeder.dev@xxxxxxxxx> writes:
>
> > See that last "no link: git-bisect" line? That's what happened to
> > catch my eyes when Derrick forgot to add his new 'sparse-checkout'
> > builtin to 'command-list.txt'. I still haven't looked up what that
> > 'no link' is supposed to mean, but if it were an error, then we would
>
> Build procedure for the Documentation uses ../command-list.txt as
> its input to produce cmds-<class>.txt files, that are included in
> git.txt (hence resulting git.1 and git.html). While it works, it
> tries to make sure that a command that has its own documentation
> page at Documentation/git-<command>.txt is listed; otherwise we have
> a page for <command> to which there is no link from git.{1,html}.
Oh, then it doesn't quite do what I thought it does.
> I think "make check-builtins" might also want to sanity check
> command-list.txt.
I haven't noticed that we have something like that. FWIW, our
documentation CI job runs it already.
And indeed it can be easily extended to check 'command-list.txt' as
well, but then it will find some other builtins not included in
'command-list.txt', namely:
bisect--helper env--helper fsck-objects init-db merge-ours
merge-recursive merge-subtree remote-ext remote-fd submodule--helper
I think it makes sense not to include the '*--helper' commands, or
'fsck-objects' and 'init-db', but I'm not sure about the others.
OTOH, it won't help if we add a new git-foo script and forgot to
include it in 'command-list.txt'.
