[Tarantool-patches] [PATCH v2 09/10] wal: fix say_x format

[Vladislav Shpilevoy]

Thanks for the patch!

On 24.02.2021 16:36, Cyrill Gorcunov via Tarantool-patches wrote:
>  - vclock_get returns int64_t
>  - vclock_sum returns int64_t
> 
> Part-of #5846
> 
> Signed-off-by: Cyrill Gorcunov <gorcunov at gmail.com>
> ---
>  src/box/wal.c | 15 ++++++++-------
>  1 file changed, 8 insertions(+), 7 deletions(-)
> 
> diff --git a/src/box/wal.c b/src/box/wal.c
> index 937d47ba9..1c0076f26 100644
> --- a/src/box/wal.c
> +++ b/src/box/wal.c
> @@ -993,13 +993,14 @@ wal_assign_lsn(struct vclock *vclock_diff, struct vclock *base,
>  			int64_t diff = (*row)->lsn - vclock_get(base, (*row)->replica_id);
>  			if (diff <= vclock_get(vclock_diff,
>  					       (*row)->replica_id)) {
> +				int64_t confirmed_lsn =
> +					vclock_get(base, (*row)->replica_id) +
> +					vclock_get(vclock_diff, (*row)->replica_id);
>  				say_crit("Attempt to write a broken LSN to WAL:"
> -					 " replica id: %d, confirmed lsn: %d,"
> -					 " new lsn %d", (*row)->replica_id,
> -					 vclock_get(base, (*row)->replica_id) +
> -					 vclock_get(vclock_diff,
> -						    (*row)->replica_id),
> -						    (*row)->lsn);
> +					 " replica id: %d, confirmed lsn: %lld,"

`replica_id` is uint32_t. So it should be %u and cast to (unsigned) I suppose.

> +					 " new lsn %lld", (*row)->replica_id,
> +					 (long long)confirmed_lsn,
> +					 (long long)(*row)->lsn);
>  				assert(false);
>  			} else {
>  				vclock_follow(vclock_diff, (*row)->replica_id, diff);
> @@ -1241,7 +1242,7 @@ wal_write_async(struct journal *journal, struct journal_entry *entry)
>  		 */
>  		say_error("Aborting transaction %llu during "
>  			  "cascading rollback",
> -			  vclock_sum(&writer->vclock));
> +			  (long long)vclock_sum(&writer->vclock));

Long long should use %lld, not %llu. I have no idea why is the format checker
silent about this.