From: Junio C Hamano <gitster@pobox.com>
Date: Wed, 08 Dec 2010 22:50:48 GMT
Subject: Re: [PATCH v2.1 3/4] describe: Store commit_names in a hash table by commit SHA1
Message-ID: <7v1v5rkhfb.fsf@alter.siamese.dyndns.org>
In-Reply-To: <alpine.DEB.2.02.1012081319320.23348@dr-wily.mit.edu>

Anders Kaseorg <andersk@ksplice.com> writes:

> @@ -44,6 +47,19 @@ static const char *prio_names[] = {
>  	"head", "lightweight", "annotated",
>  };
>  
> +static inline unsigned int hash_sha1(const unsigned char *sha1)
> +{
> +	return *(unsigned int *)sha1;
> +}

Do we know that all the archs we will be compiled on will be happy with
this potentially unaligned access?  hash_filespec() in diffcore-rename.c
is written in a way to avoid such an issue, and I would feel safer to see
this do the same.