We have set up a new test environment running PostgreSQL v12.2 on macOS
10.14 and the issue is still there.
One thing I noticed, is that the returning columns do not affect the
behaviour:
SELECT * FROM f_gsxws_schedule WHERE UPPER(gwsc_dossier) = 'TEST'
and
SELECT gwsc_sequence FROM f_gsxws_schedule WHERE UPPER(gwsc_dossier) =
'TEST'
both kill the postgres service.
I will try to free some time next week to install the Apple developer
tools and further analyse the problem.
Best regards,
Nick
On 11 Feb 2020, at 12:32, Nick Renders wrote:
Hi Thomas,
We are setting up a new test environment with 12.1.
Once it is running, I'll try out those commands and get back with the
results.
Thanks,
Nick Renders
On 11 Feb 2020, at 2:51, Thomas Munro wrote:
On Mon, Feb 10, 2020 at 4:35 AM Marc <postgres@xxxxxxxxxx> wrote:
We will keep the 12.1 in place so that we can run additional tests
to assist to pin-point the issue.
Feel free to ask but allow us to recover from these hectic days ;-)
Here's how to get a stack so we can see what it was doing, assuming
you have the Apple developer tools installed:
1. Find the PID of the backend you're connected to with SELECT
pg_backend_pid().
2. "lldb -p PID" from a shell to attach to the process, then "cont"
to let it continue running.
3. Run the query in that backend and wait for the SIGKILL.
4. In the lldb session, type "bt".
It'll only make sense if your PostgreSQL build has debug symbols, but
let's see.
