Re: TimeZone issue with 42.2.1 - Mailing list pgsql-jdbc

From Bear Giles
Subject Re: TimeZone issue with 42.2.1
Date
Msg-id CALBNtw4zejOnhKt3zZiD=uurGsrUThf8X1B8RN1hudMPDQKRJw@mail.gmail.com
Whole thread Raw
In response to Re: TimeZone issue with 42.2.1  (Vladimir Sitnikov <sitnikov.vladimir@gmail.com>)
List pgsql-jdbc
Thanks!

This explanation goes a long way towards explaining a different issue we've had with inserting data into arrays. We use JSON for internal communications and have been seeing some weirdness with it as well. The details are different but it's another place where we're using Jackson to serialize and deserialize something. I've seen messages with both "unable to convert java.util.ArrayList to java.sql.Array" and references to PgArray.

Bear

On Mon, Mar 5, 2018 at 3:23 PM, Vladimir Sitnikov <sitnikov.vladimir@gmail.com> wrote:
>an not resolve type id for sun.util.calendar.ZoneInfo

This sounds like a Jackson error.

It looks like you are trying to Jackson-serialize the result of col2 column.
The column is <<time with time zone[]>>, so it is represented as PgArray object.

As Jackson serializes PgArray, it treats whatever getter it finds as a property.
It runs into PgArray#getResultSet and thinks it is a getter for resultSet property. Of course that causes it to serialize underlying statement, connection, etc.
Finally it ends up with serializing QueryExecutorImpl.timeZone field, and it fails so.

One of the recommendations might include "avoid Jackson-serializing random objects" =)
You might have better luck if you serialize PgArray manually or provide a Jackson mixin annotation (see https://github.com/FasterXML/jackson-docs/wiki/JacksonMixInAnnotations ). Technically speaking, PgArray#getArray might be just enough.

You might ask "why did it work before 42.2.1?", and it turns out the trigger line is https://github.com/pgjdbc/pgjdbc/commit/a94cfeace5d66b4fe8d8fa3b16986baebaec2a11#diff-0ac94407005dd81cdbe1efa6066de013R2624 
The thing is backend sends connection timezone at connection setup phase, and pgjdbc 42.2.0 ignored that. pgjdbc 42.2.1 parses timezone message in the initial parameter status messages, thus it ends up with non-null timezone, thus it fails your app.


Vladimir

pgsql-jdbc by date:

Previous
From: Vladimir Sitnikov
Date:
Subject: Re: TimeZone issue with 42.2.1
Next
From: jwhiting@redhat.com
Date:
Subject: Re: Handling transaction failure due to concurrency errors