Re: [PATCH] statinfo.h: move DTYPE defines from dir.h

From: Junio C Hamano <gitster@pobox.com>
Date: Jun 3, 2023, 01:56 UTC
Message-ID: <xmqqilc571hf.fsf@gitster.g>
In-Reply-To: <20230602192700.1548636-1-asedeno@google.com>

"Alejandro R Sedeño" <asedeno@google.com> writes:

Show 13 quoted lines

> From: Alejandro R. Sedeño <asedeno@mit.edu>
>
> These definitions are used in cache.h, which can't include dir.h
> without causing name-hash.c to have two definitions of
> `struct dir_entry`.
>
> Both dir.h and cache.h include statinfo.h, and this seems a reasonable
> place for these definitions.
>
> This change fixes a broken build issue on old SunOS.
>
> Signed-off-by: Alejandro R. Sedeño <asedeno@mit.edu>
> Signed-off-by: Alejandro R Sedeño <asedeno@google.com>

This is a bit unusual; do you want to publish both names (I am assuming that they are the same single person)?

I thought somebody in the earlier discussion identified the topic that was problematic by bisecting. It is a shame to lose that. Perhaps it is a good idea to rephrase the beginning of the proposed commit log message to mention that, like

    592fc5b3 (dir.h: move DTYPE defines from cache.h, 2023-04-22)
    moved DTYPE macros from cache.h to dir.h, but are still used
    by cache.h to implement ce_to_dtype(); but cache.h cannot
    include dir.h because ...

or something?

Why does name-hash.c end up with two definitions? Aren't we properly guarding against multiple inclusions with

    #ifndef __DIR_H__
    #define __DIR_H__
	...
    struct dir_entry {
	...
    };
    #endif

or is there something funny going on?

Thanks.