docs: document zero bits in index "mode"

Commit Message

Glen Choo Feb. 1, 2023, 2:40 a.m. UTC

Documentation/gitformat-index.txt describes the "mode" as 32 bits, but
only documents 16 bits. Document the missing 16 bits and specify that
'unused' bits must be zero.

Signed-off-by: Glen Choo <[e-mail]>
---
>                 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.

If I am understanding you correctly (which I'm not sure, since I am
honestly clueless about bit numbering and big endianness and whatnot),
you're saying that highest 16 bits are zero, not the lowest? If so, then
I genuinely made that mistake, hah.

(We're storing 16 bits as bigendian 32 bits, so they would occupy the
lower 16 bits, right..?)

Perhaps something like this would be better. I took the "unused, must be
zero" phrasing from elsewhere in the doc. And if I am completely
off-base, feel free to patch it without waiting for my reroll if you
think that's easier.

 Documentation/gitformat-index.txt | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

Comments

Glen Choo Feb. 1, 2023, 2:46 a.m. UTC | #1

Glen Choo <[e-mail]> writes:

>>                 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.
>
> If I am understanding you correctly (which I'm not sure, since I am
> honestly clueless about bit numbering and big endianness and whatnot),
> you're saying that highest 16 bits are zero, not the lowest? If so, then
> I genuinely made that mistake, hah.
>
> (We're storing 16 bits as bigendian 32 bits, so they would occupy the
> lower 16 bits, right..?)
>
> Perhaps something like this would be better. I took the "unused, must be
> zero" phrasing from elsewhere in the doc. And if I am completely
> off-base, feel free to patch it without waiting for my reroll if you
> think that's easier.

Whoops, this was meant to be a reply to
https://lore.kernel.org/git/xmqqmt5yy08d.fsf@gitster.g. That's what I
get for trying to write the In-Reply-To by hand.

Junio C Hamano Feb. 1, 2023, 4:56 p.m. UTC | #2

Glen Choo <[e-mail]> writes:

> Documentation/gitformat-index.txt describes the "mode" as 32 bits, but
> only documents 16 bits. Document the missing 16 bits and specify that
> 'unused' bits must be zero.

I actually think the bottom 16-bit should just be described as
allowing only a limited set of possible values and list them all,
but this is good as an incremental improvement [*]

Thanks.


>    32-bit mode, split into (high to low bits)
>  
> +    16-bit unused, must be zero
> +
>      4-bit object type
>        valid values in binary are 1000 (regular file), 1010 (symbolic link)
>        and 1110 (gitlink)
>  
> -    3-bit unused
> +    3-bit unused, must be zero
>  
>      9-bit unix permission. Only 0755 and 0644 are valid for regular files.
>      Symbolic links and gitlinks have value 0 in this field.


[Footnote]

If I am not mistaken, there are only four possible values allowed.
So, instead of "split into...", we could just say

        32-bit "mode", which can be one of

          Regular non-executable file     100644
          Regular executable file         100755
          Symbolic link                   120000
          Git link                        160000

        No other values are allowed.

but we can move one step at a time ;-)

Patch

diff --git a/Documentation/gitformat-index.txt b/Documentation/gitformat-index.txt
index 015cb21bdc..0773e5c380 100644
--- a/Documentation/gitformat-index.txt
+++ b/Documentation/gitformat-index.txt
@@ -83,11 +83,13 @@  Git index format
 
   32-bit mode, split into (high to low bits)
 
+    16-bit unused, must be zero
+
     4-bit object type
       valid values in binary are 1000 (regular file), 1010 (symbolic link)
       and 1110 (gitlink)
 
-    3-bit unused
+    3-bit unused, must be zero
 
     9-bit unix permission. Only 0755 and 0644 are valid for regular files.
     Symbolic links and gitlinks have value 0 in this field.