gitformat-index.txt has a gap in the "mode" description?

Glen Choo· Jan 31, 2023, 16:43 UTC · lore

According to gitformat-index.txt [1], "mode" is 32 bits, but we've only documented 16 bits. I tried poking around read-cache.c, and my impression is that other 16 are just NUL. If so, it would be worth documenting that they're unused, especially since we documented unused bits right in that section.

[1] https://github.com/git/git/blob/master/Documentation/gitformat-index.txt#L84

Junio C Hamano· Jan 31, 2023, 19:26 UTC · re: Glen Choo · lore

Re: gitformat-index.txt has a gap in the "mode" description?

Glen Choo <chooglen@google.com> writes:

Show 7 quoted lines

> According to gitformat-index.txt [1], "mode" is 32 bits, but we've only
> documented 16 bits. I tried poking around read-cache.c, and my
> impression is that other 16 are just NUL. If so, it would be worth
> documenting that they're unused, especially since we documented
> unused bits right in that section.
>
> [1] https://github.com/git/git/blob/master/Documentation/gitformat-index.txt#L84

Good thinking. The existing explanation starts with "32-bit mode, split into (high to low bits)", followed by "4-bit object type", as if the "4-bit object type" occupies bits 29-32, which is not quite what we want to say.

By the way, we should avoid saying "unused" but be more explicit and say "must be zero". The latter has no ambiguity but the former can be misinterpreted to allow any garbage.

Thanks.