On Sunday, June 26, 2022, WR <
wolle321@freenet.de> wrote:
I made some test with pgadmin. Pgadmin (5.2) also reports this error now. And it doesn't matter if standard_conforming_strings is on or off.
SET standard_conforming_strings = off;
INSERT INTO public.oned_figures VALUES (1, 'Figure_Wolle1', 476, -476, 2000, 2400, 2400, '\000', 500, 0, 'sinus(0|0|0;30;5;0;0,5;0)', '2021-08-31 11:53:22.442801', 0, 1);
And
SET standard_conforming_strings = on;
INSERT INTO public.oned_figures VALUES (1, 'Figure_Wolle1', 476, -476, 2000, 2400, 2400, '\000', 500, 0, 'sinus(0|0|0;30;5;0;0,5;0)', '2021-08-31 11:53:22.442801', 0, 1);
Both do report:
WARNUNG: nicht standardkonforme Verwendung von Escape in Zeichenkettenkonstante
LINE 8: ...(1, 'Figure_Wolle1', 476, -476, 2000, 2400, 2400, '\000\000&...
^
HINT: Verwenden Sie die Syntax für Escape-Zeichenketten, z.B. E'\r\n'.
ERROR: FEHLER: ungültige Byte-Sequenz für Kodierung »UTF8«: 0x00
I’m not in a position to test/experiment on any guaranteed timeframe but your observation that the outcome of those two commands is independent of value of standard_conforming_strings is either a straight up bug or you’ve made a mistake somewhere. The warning is fully conditioned on that setting being off.
E’\\000’ and ‘\000’ passed as string literals to the bytea input routine are defined to be equivalent inputs under standard_conforming_strings and neither can produce the warning in that case.
I suggest doing self-contained examples that demonstrate the documented behavior not working as documented (or not being functional even if intended) to pinpoint any bug that might be lurking here. With only fragments and statements that seem impossible we are left to assume operator error. pg_dump is completely correct in what it is producing (non-escape literal \000).
I also suggest using psql and pg_dump directly, and not pgAdmin, to demonstrate a core PostgreSQL bug.
David J.