On Tue, 22 Jun 2010, S K wrote:
>
> Our application has been in production for a couple of years. Recent
> investigation into a problem revealed the following error started a few
> months back. I restarted the application, but the error appears after a few
> days and the thread dies.
>
> Caused by: java.lang.NegativeArraySizeException
> at org.postgresql.core.PGStream.ReceiveTupleV3(PGStream.java:349)
> atorg.postgresql.core.v3.QueryExecutorImpl.processResults(QueryExecutorImpl.j
> ava:1306)
> atorg.postgresql.core.v3.QueryExecutorImpl.execute(QueryExecutorImpl.java:192
> )
> atorg.postgresql.jdbc2.AbstractJdbc2Statement.execute(AbstractJdbc2Statement.
> java:451)
> atorg.postgresql.jdbc2.AbstractJdbc2Statement.executeWithFlags(AbstractJdbc2S
> tatement.java:350)
> atorg.postgresql.jdbc2.AbstractJdbc2Statement.executeQuery(AbstractJdbc2State
> ment.java:254)
I haven't seen an error like this before. It most likely indicates that
the driver has lost protocol sync with the server and is reading the wrong
part of the message. This is difficult to debug because it's unknown how
far back synchronization was lost. To try to debug this I would modify
the driver to log the negative array size value and as much of the message
parts I could find around it. That may give some clues as to why the
driver is confused. Additionally, I would try to determine the triggering
query and see if I could reproduce it.
Kris Jurka
