> On 26 Feb 2019, at 18:36, Imeev Mergen <imeevma@tarantool.org> wrote:
> Hi! Thank you for review! My answers below.
> On 2/26/19 5:47 PM, n.pettik wrote:
>>> diff --git a/src/box/sql.c b/src/box/sql.c
>>> index 580f3fa..116e3e8 100644
>>> --- a/src/box/sql.c
>>> +++ b/src/box/sql.c
>>> @@ -1362,13 +1362,8 @@ sql_checks_resolve_space_def_reference(ExprList *expr_list,
>>> 	parser.parse_only = true;
>>> 
>>> 	sql_resolve_self_reference(&parser, def, NC_IsCheck, NULL, expr_list);
>>> -	int rc = 0;
>>> -	if (parser.rc != SQL_OK) {
>>> -		/* Tarantool error may be already set with diag. */
>>> -		if (parser.rc != SQL_TARANTOOL_ERROR)
>>> -			diag_set(ClientError, ER_SQL, parser.zErrMsg);
>>> -		rc = -1;
>>> -	}
>>> +	if (parser.rc != SQL_OK)
>>> +		return -1;
>> Since now we have only one possible RC, lets remove
>> its name and simply check (parser.rc != 0).
>> Or, as suggested Konstantin, better replace rc with bool is_aborted flag.
> Do you mind if I do this in a new patch?

Sure, but place it before this one.

>>> diff --git a/src/box/sql/build.c b/src/box/sql/build.c
>>> index deb5b89..6afca4a 100644
>>> --- a/src/box/sql/build.c
>>> +++ b/src/box/sql/build.c
>>> @@ -493,16 +493,10 @@ sql_column_add_nullable_action(struct Parse *parser,
>>> 	struct field_def *field = &def->fields[def->field_count - 1];
>>> 	if (field->nullable_action != ON_CONFLICT_ACTION_DEFAULT &&
>>> 	    nullable_action != field->nullable_action) {
>>> -		/* Prevent defining nullable_action many times. */
>>> -		const char *err_msg =
>>> -			tt_sprintf("NULL declaration for column '%s' of table "
>>> -				   "'%s' has been already set to '%s'",
>>> -				   field->name, def->name,
>>> -				   on_conflict_action_strs[field->
>>> -							   nullable_action]);
>>> -		diag_set(ClientError, ER_SQL, err_msg);
>>> -		parser->rc = SQL_TARANTOOL_ERROR;
>>> -		parser->nErr++;
>>> +		sqlErrorMsg(parser, "NULL declaration for column '%s' of "\
>>> +			    "table '%s' has been already set to '%s'",
>>> +			    field->name, def->name,
>>> +			    on_conflict_action_strs[field-> nullable_action]);
>> This looks like step back in our attempt at using diag_set.
>> We do you need to incapsulate diag into sqlErrorMsg?
> I thought that it was good idea to incapsulate all SQL errors that

No, it wasn’t.

> cannot be fixed in just parse_context->rc = SQL_TARANTOOL_ERROR
> and diag_set(). Here it uses tt_printf() in addition to these two
> commands.

It’s OK. Revert this (and other) change(s), please. Lets use pure
diag_set + abort flag.

>>> 		return;
>>> 	}
>>> 	field->nullable_action = nullable_action;
>>> diff --git a/src/box/sql/delete.c b/src/box/sql/delete.c
>>> index 5170c7f..a7bf3b3 100644
>>> --- a/src/box/sql/delete.c
>>> +++ b/src/box/sql/delete.c
>>> @@ -94,31 +94,22 @@ sql_table_truncate(struct Parse *parse, struct SrcList *tab_list)
>>> 	struct space *space = space_by_name(tab_name);
>>> 	if (space == NULL) {
>>> 		diag_set(ClientError, ER_NO_SUCH_SPACE, tab_name);
>>> -		goto tarantool_error;
>>> +		sql_parser_error(parse);
>> Look, anyway you remove this function in next commit.
>> Next time please consider order of refactoring.
> You are right, I will return to what was before and refactor in
> the next commit.

Don’t waste time now, it was just advice to keep in mind.

>>> 	}
>>> 	if (! rlist_empty(&space->parent_fk_constraint)) {
>>> -		const char *err_msg =
>>> -			tt_sprintf("can not truncate space '%s' because other "
>>> -				   "objects depend on it", space->def->name);
>>> -		diag_set(ClientError, ER_SQL, err_msg);
>>> -		goto tarantool_error;
>>> +		sqlErrorMsg(parse, "can not truncate space '%s' because other "
>>> +			    "objects depend on it", space->def->name);
>> Replace invocation of sqlErrorMsg with diag_set + parser->rc.
>> The same in other places.
> Answer is the same as in second question.
>> 
>>> @@ -146,11 +145,7 @@ sqlPrepare(sql * db,	/* Database handle. */
>>> 		*ppStmt = (sql_stmt *) sParse.pVdbe;
>>> 	}
>>> 
>>> -	if (zErrMsg) {
>>> -		sqlErrorWithMsg(db, rc, "%s", zErrMsg);
>>> -	} else {
>>> -		sqlError(db, rc);
>>> -	}
>>> +	sqlError(db, rc);
>> sqlError seems to be useless/dead. Please, make a note somewhere
>> to remove it as follow-up to error-refactoring patch-set.
> Do you mind if I do this in a new patch?

You don’t have to do it right now, it can wait till 2.2
Just file this issue somewhere.