Re: [HACKERS] WIP Patch: Pgbench Serialization and deadlock errors

On 13-06-2018 22:59, Alvaro Herrera wrote:
> For context: in the backend, elog() is only used for internal messages
> (i.e. "can't-happen" conditions), and ereport() is used for user-facing
> messages. There are many things ereport() has that elog() doesn't,
> such
> as additional message fields (HINT, DETAIL, etc) that I think could
> have
> some use in pgbench as well. If you use elog() then you can't have
> that.

AFAIU originally it was not supposed that the pgbench error messages
have these fields, so will it be good to change the final output to
stderr?.. For example:

- fprintf(stderr, "%s", PQerrorMessage(con));
- fprintf(stderr, "(ignoring this error and continuing anyway)\n");
+ ereport(LOG,
+ (errmsg("Ignoring the server error and continuing anyway"),
+ errdetail("%s", PQerrorMessage(con))));

- fprintf(stderr, "%s", PQerrorMessage(con));
- if (sqlState && strcmp(sqlState, ERRCODE_UNDEFINED_TABLE) == 0)
- {
- fprintf(stderr, "Perhaps you need to do initialization (\"pgbench
-i\") in database \"%s\"\n", PQdb(con));
- }
-
- exit(1);
+ ereport(ERROR,
+ (errmsg("Server error"),
+ errdetail("%s", PQerrorMessage(con)),
+ sqlState && strcmp(sqlState, ERRCODE_UNDEFINED_TABLE) == 0 ?
+ errhint("Perhaps you need to do initialization (\"pgbench -i\")
in database \"%s\"\n",
+ PQdb(con)) : 0));

--
Marina Polyakova
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company