Re: [RFC PATCH 5/6] sha1-name: Teach `get_oid_with_context[_1]()` to be gentle

From: Junio C Hamano <gitster@pobox.com>
Date: Jul 17, 2018, 19:13 UTC
Message-ID: <xmqqefg1oeuk.fsf@gitster-ct.c.googlers.com>
In-Reply-To: <dc1abfca46cd536f060b9097b9f2d1a64988f932.1531778417.git.ungureanupaulsebastian@gmail.com>

Paul-Sebastian Ungureanu <ungureanupaulsebastian@gmail.com> writes:

Show 10 quoted lines

> @@ -1769,7 +1831,8 @@ void maybe_die_on_misspelt_object_name(const char *name, const char *prefix)
>  
>  int get_oid_with_context(const char *str, unsigned flags, struct object_id *oid, struct object_context *oc)
>  {
> -	if (flags & GET_OID_FOLLOW_SYMLINKS && flags & GET_OID_ONLY_TO_DIE)
> +	if (flags & (GET_OID_FOLLOW_SYMLINKS | GET_OID_GENTLY) &&
> +	    flags & GET_OID_ONLY_TO_DIE)
>  		BUG("incompatible flags for get_sha1_with_context");
>  	return get_oid_with_context_1(str, flags, NULL, oid, oc);
>  }

This points us back to "only-to-die" which was "gently" before 2e83b66c ("fix overslow :/no-such-string-ever-existed diagnostics", 2011-05-10). I think we have to keep them both, as only-to-die means more than just being not gentle, and we cannot revert the renaming s/!gently/only-to-die/ done by 2e83b66c and teach GENTLY to more codepaths, I think. But I might be mistaken and we may be able to get rid of only-to-die at the end of this series. I dunno.

In any case, what's the reason why this new "gentle" option is incompatible with "only-to-die"?