Re: [Tarantool-patches] [PATCH 05/17] recovery: recover_xlog -- don't throw exception

[Vladislav Shpilevoy <v.shpilevoy@tarantool.org>]

Thanks for the patch!

See 3 comments below.

> diff --git a/src/box/recovery.cc b/src/box/recovery.cc
> index 55d89903f..f724600ed 100644
> --- a/src/box/recovery.cc
> +++ b/src/box/recovery.cc
> @@ -243,25 +243,28 @@ recovery_delete(struct recovery *r)
>   * The reading will be stopped on reaching stop_vclock.
>   * Use NULL for boundless recover
>   */
> -static void
> +static int
>  recover_xlog(struct recovery *r, struct xstream *stream,
>  	     const struct vclock *stop_vclock)
>  {
> +	bool force_recovery = r->wal_dir.force_recovery;
>  	struct xrow_header row;
>  	uint64_t row_count = 0;
> -	while (xlog_cursor_next_xc(&r->cursor, &row,
> -				   r->wal_dir.force_recovery) == 0) {

1. This function becomes unused and can be deleted now.

> +	int rc;
> +
> +	while ((rc = xlog_cursor_next(&r->cursor, &row, force_recovery)) == 0) {
> @@ -272,6 +275,7 @@ recover_xlog(struct recovery *r, struct xstream *stream,
>  		 * are signed with a zero replica id.
>  		 */
>  		assert(row.replica_id != 0 || row.group_id == GROUP_LOCAL);
> +

2. Lets better omit unnecessary diff, even if it sometimes looks like
it makes some parts of the code better. At least when it increases
number of diff hunks.

>  		/*
>  		 * We can promote the vclock either before or
>  		 * after xstream_write(): it only makes any impact
> @@ -281,18 +285,24 @@ recover_xlog(struct recovery *r, struct xstream *stream,
>  		vclock_follow_xrow(&r->vclock, &row);
>  		if (xstream_write(stream, &row) == 0) {
>  			++row_count;
> -			if (row_count % 100000 == 0)
> +			if (row_count % 100000 == 0) {
>  				say_info("%.1fM rows processed",
>  					 row_count / 1000000.);
> -		} else {
> -			if (!r->wal_dir.force_recovery)
> -				diag_raise();
> +			}
> +			continue;
> +		}
>  
> -			say_error("skipping row {%u: %lld}",
> -				  (unsigned)row.replica_id, (long long)row.lsn);
> -			diag_log();
> +		if (!force_recovery) {
> +			rc = -1;
> +			break;
>  		}
> +
> +		say_error("skipping row {%u: %lld}",
> +			  (unsigned)row.replica_id, (long long)row.lsn);
> +		diag_log();
>  	}
> +
> +	return rc;

3. Probably better return 0/-1 just like all the other functions, which
can either fail or succeed. Otherwise some new code in future for sure
will check result using '!= 0', and that will be a bug. Because you
can return 1 now.