Can you provide the table structure and a demo for replay

Can you provide the table structure and a demo for replay

users: table

• columns
  id: bigint NN default nextval('users_id_seq'::regclass)
  . references
  sequence = users_id_seq
  name: varchar(128) default NULL::character varying
  -- Nickname
  pwd: varchar(124) default NULL::character varying
  -- Password
  head_img: varchar(524) default NULL::character varying
  -- Profile Picture
  slogan: varchar(524) default NULL::character varying
  -- User Signature
  sex: smallint default 1
  -- 0 for female, 1 for male
  points: integer default 0
  -- Points
  create_time: timestamp
  mail: varchar(64) default NULL::character varying
  -- Email
  secret: varchar(12) default NULL::character varying
  -- Salt, used for personal sensitive information processing
• indices
  #1 : unique (id)
  #2 : unique (mail)
• keys
  #1 : PK (id) (underlying index #1 )
  #2 : AK (mail) (underlying index #2 )

Can you upload a demo or use a pseudo-code description,I want to see mapper.xml how to write and some config, so can to replay.

Can you upload a demo or use a pseudo-code description,I want to see mapper.xml how to write and some config, so can to replay.

You can reference to the code here

If I add the global transactional decorator at the register function, then what's on line 87 will log a null id

had readed your case, module user-service dependence exclude not ok, actual seata version is 1.1.0. after update, it will be work will.

so why seata 1.1 result difference between mysql and postgrel when return pk to object?

when happens insert, there will be generate undolog in seata, the process will get the pk, when finish read the resultSet , mysql cursor can move back, but postgrel cursor can't.

when mybatis fill pk to object, it will get pk too, so mysql cursor can continue to read, but postgrel already finish read cursor, so can't read.

current seata version fix the issue, when generate undolog, read the pk resultset, there will be cache for pk, so next operation will read the cache.

had readed your case, module user-service dependence exclude not ok, actual seata version is 1.1.0. after update, it will be work will.

so why seata 1.1 result difference between mysql and postgrel when return pk to object?

when happens insert, there will be generate undolog in seata, the process will get the pk, when finish read the resultSet , mysql cursor can move back, but postgrel cursor can't.

when mybatis fill pk to object, it will get pk too, so mysql cursor can continue to read, but postgrel already finish read cursor, so can't read.

current seata version fix the issue, when generate undolog, read the pk resultset, there will be cache for pk, so next operation will read the cache.

I specified the version of seats in the common package to 1.7.0 which matches the seata version that I am using. I am also assume that given 1.7.0 is later that 1.1. It should have contained the fix that is implemented right?

yeah, 1.7.0 contain the fix. currently, actual seata version is 1.1.0 in your user-service module, you can explicit add seata dependency to 1.7.0 in it, have a try.