Subject: Re: PL/Java 1.5.2 - fixes one regression in date conversion in 1.5.1
To: Christoph Berg <myon@debian.org>, pgsql-pkg-debian@postgresql.org,
 pgsql-pkg-yum@postgresql.org
References: <5BE08D74.2060207@anastigmatix.net>
 <20181106074931.GB3898@msg.df7cb.de> <5BE18B2F.9030805@anastigmatix.net>
 <20181106145659.GC13511@msg.df7cb.de>
From: Chapman Flack <chap@anastigmatix.net>
Message-ID: <5BE32DAE.7090508@anastigmatix.net>
Date: Wed, 7 Nov 2018 13:23:42 -0500
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
 Thunderbird/38.7.0
MIME-Version: 1.0
In-Reply-To: <20181106145659.GC13511@msg.df7cb.de>
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 11/06/18 09:56, Christoph Berg wrote:

> There's still a problem, though. Jessie, 9.3, OpenJDK 7:
> 
> ERROR:  java.lang.SecurityException: read on /usr/share/javazi/Europe/Prague
> STATEMENT:  SELECT sqlj.install_jar('file:/usr/share/postgresql/9.3/pljava/pljava-examples-1.5.2.jar', 'examples', true)

Hmm. Does that happen only on Jessie? Is there a debianism where zoneinfo
files for Java are distributed in this other location instead of under
$JAVA_HOME ?

At present, the security manager for "trusted" PL/Java doesn't offer any
easy way to configure additional filesystem locations that are ok to read
from. I'm working on that for a future version, but that doesn't help here.

I suspect it could be made to work by adding
trust=org.postgresql.pljava.annotation.Function.Trust.UNSANDBOXED
in the @Function annotation on the issue199() method in
pljava-examples/src/main/java/org/postgresql/pljava/example/annotation/PreJSR310.java

However, it is only a regression test, and it is probably simpler just to
build with a patch to not run it (say, by removing the @SQLAction annotation
at the top of that file). That's preferable to adding UNSANDBOXED willy
nilly to things in the examples jar that people may casually install.

This is good for me to know about, as I was thinking of changing PL/Java's
behavior in a future release to set Java's timezone to match the PG
session's by default. Now I know not to do that until there is a way to
accommodate zoneinfo from atypical locations.

-Chap