Re: [PATCH 2/2] refactor name_from_dns

[Rich Felker <dalias@libc.org>]

On Wed, May 25, 2016 at 11:22:14AM +0200, Natanael Copa wrote:
> loop over an address family / resource record mapping to avoid
> repetitive code.
> ---
>  src/network/lookup_name.c | 27 +++++++++++++--------------
>  1 file changed, 13 insertions(+), 14 deletions(-)
> 
> diff --git a/src/network/lookup_name.c b/src/network/lookup_name.c
> index d3d97b4..99364ab 100644
> --- a/src/network/lookup_name.c
> +++ b/src/network/lookup_name.c
> @@ -141,20 +141,19 @@ static int name_from_dns(struct address buf[static MAXADDRS], char canon[static
>  	int qlens[2], alens[2];
>  	int i, nq = 0;
>  	struct dpc_ctx ctx = { .addrs = buf, .canon = canon };
> -
> -	if (family != AF_INET6) {
> -		qlens[nq] = __res_mkquery(0, name, 1, RR_A, 0, 0, 0,
> -			qbuf[nq], sizeof *qbuf);
> -		if (qlens[nq] == -1)
> -			return EAI_NONAME;
> -		nq++;
> -	}
> -	if (family != AF_INET) {
> -		qlens[nq] = __res_mkquery(0, name, 1, RR_AAAA, 0, 0, 0,
> -			qbuf[nq], sizeof *qbuf);
> -		if (qlens[nq] == -1)
> -			return EAI_NONAME;
> -		nq++;
> +	struct { int af; int rr; } afrr[2] = {
> +		{ .af = AF_INET6, .rr = RR_A },
> +		{ .af = AF_INET, .rr = RR_AAAA },
> +	};
> +
> +	for (i=0; i<2; i++) {
> +		if (family != afrr[i].af) {
> +			qlens[nq] = __res_mkquery(0, name, 1, afrr[i].rr,
> +				0, 0, 0, qbuf[nq], sizeof *qbuf);
> +			if (qlens[nq] == -1)
> +				return EAI_NONAME;
> +			nq++;
> +		}

Can you compare the codegen for this as written vs making the table
static-const? I wonder which gcc does better with. My guess would be
static-const is better if it actually emits the table and loops over
it, but auto storage like you did might be better if that helps it
just unroll it plugging in the values.

Rich