January 2026 - Hackers - lists.ivorysql.org

Re: When a query returns an empty result set ...

by Oreo Yang

Hi Pierre, > > I would like to know if this is intended behavior ? > We have successfully reproduced the issue and can confirm the behavioral difference you've observed. Yes, that is the expected behavior. Root Cause Analysis: This difference occurs because IvorySQL's PL/iSQL currently follows PostgreSQL's PL/pgSQL semantics rather than Oracle's. ---- Thanks, Oreo

Jan. 30, 2026

When a query returns an empty result set ...

by Pierre Forstmann

... In Oracle PL/SQL, an error 1403 is raised: SQL> -- SQL> drop table t; Table dropped. SQL> create table t(dummy int); Table created. SQL> declare 2 x varchar2(1); 3 -- l text; 4 l varchar2(40); 5 status integer; 6 high integer := 10; 7 begin 8 dbms_output.enable(); 9 for i in 1..high loop 10 select dummy into x from t; 11 dbms_output.put_line(to_char(i)); 12 end loop; 13 for i in 1..high loop 14 dbms_output.get_line(l, status); 15 -- raise notice 'i=% l=%', i, l; 16 end loop; 17 end; 18 / declare * ERROR at line 1: ORA-01403: no data found ORA-06512: at line 10 Help: https://docs.oracle.com/error-help/db/ora-01403/ ... In IvorySQL PL/ISQL, no error has raised and the target output variable is NULL (like in PL/PGSQL): $ cat t1.sql drop table t; create table t(dummy int); -- PLI/PGSQL declare x varchar2(1); l text; status integer; high integer := 10; begin dbms_output.enable(); for i in 1..high loop select dummy into x from t; dbms_output.put_line(to_char(i)); end loop; for i in 1..high loop dbms_output.get_line(l, status); raise notice 'i=% l=%', i, l; end loop; end; / $ psql -p 1521 -f t1.sql DROP TABLE CREATE TABLE psql:t1.sql:22: NOTICE: i=1 l=1 psql:t1.sql:22: NOTICE: i=2 l=2 psql:t1.sql:22: NOTICE: i=3 l=3 psql:t1.sql:22: NOTICE: i=4 l=4 psql:t1.sql:22: NOTICE: i=5 l=5 psql:t1.sql:22: NOTICE: i=6 l=6 psql:t1.sql:22: NOTICE: i=7 l=7 psql:t1.sql:22: NOTICE: i=8 l=8 psql:t1.sql:22: NOTICE: i=9 l=9 psql:t1.sql:22: NOTICE: i=10 l=10 DO $ I would like to know if this is intended behavior ? Thanks.

Jan. 29, 2026

Re: Error message wording

by Pierre Forstmann

Hi, I have created PR: https://github.com/IvorySQL/IvorySQL/pull/1172 Thanks and regards.

Jan. 24, 2026

Re: make installcheck failures

by Pierre Forstmann

Hi, Yes it has helped: I was able to run "make installcheck" without errors. I have closed the issue. Thanks and regards. Pierre

Jan. 24, 2026

Re: make installcheck failures

by jiaoshuntian＠highgo.com

Hi Pierre, I replied to your confusion in issues/1171, hope it helps. see https://github.com/IvorySQL/IvorySQL/issues/1171 Regards, JiaoShuntian

Jan. 24, 2026

Re: Error message wording

by jiaoshuntian＠highgo.com

Hi, Pierre I agree with you. The comments in your patch are better than the current description. Looking forward to you contributing it to the community repository. Thanks

Jan. 24, 2026

make installcheck failures

by Pierre Forstmann

Hello, I have built IvorySQL from source and I try to run "install makecheck". It fails on: make[5]: Entering directory '/home/pierre/IvorySQL/src/common' make[5]: Nothing to be done for 'all'. make[5]: Leaving directory '/home/pierre/IvorySQL/src/common' make[4]: Leaving directory '/home/pierre/IvorySQL/src/test/regress' echo "# +++ regress install-check in src/pl/plpgsql/src +++" && ../../../../src/test/regress/pg_regress --inputdir=. --bindir='/usr/local/ivorysql/ivorysql-5/bin' --dbname=pl_regression plpgsql_array plpgsql_cache plpgsql_call plpgsql_control plpgsql_copy plpgsql_domain plpgsql_misc plpgsql_record plpgsql_simple plpgsql_transaction plpgsql_trap plpgsql_trigger plpgsql_varprops # +++ regress install-check in src/pl/plpgsql/src +++ # using postmaster on Unix socket, default port not ok 1 - plpgsql_array 24 ms ok 2 - plpgsql_cache 13 ms not ok 3 - plpgsql_call 52 ms not ok 4 - plpgsql_control 47 ms ok 5 - plpgsql_copy 30 ms ok 6 - plpgsql_domain 48 ms ok 7 - plpgsql_misc 11 ms not ok 8 - plpgsql_record 117 ms ok 9 - plpgsql_simple 25 ms ok 10 - plpgsql_transaction 109 ms ok 11 - plpgsql_trap 1060 ms ok 12 - plpgsql_trigger 13 ms not ok 13 - plpgsql_varprops 15 ms 1..13 # 5 of 13 tests failed. # The differences that caused some tests to fail can be viewed in the file "/home/pierre/IvorySQL/src/pl/plpgsql/src/regression.diffs". # A copy of the test summary that you see above is saved in the file "/home/pierre/IvorySQL/src/pl/plpgsql/src/regression.out". make[3]: *** [Makefile:101: installcheck] Error 1 make[3]: Leaving directory '/home/pierre/IvorySQL/src/pl/plpgsql/src' make[2]: *** [Makefile:17: installcheck-src-recurse] Error 2 make[2]: Leaving directory '/home/pierre/IvorySQL/src/pl/plpgsql' make[1]: *** [Makefile:36: installcheck-plpgsql-recurse] Error 2 make[1]: Leaving directory '/home/pierre/IvorySQL/src/pl' make: *** [Makefile:48: installcheck-pl-recurse] Error 2 Example of edtailed error messages: diff -U3 /home/pierre/IvorySQL/src/pl/plpgsql/src/expected/plpgsql_array.out /home/pierre/IvorySQL/src/pl/plpgsql/src/results/plpgsql_array.out --- /home/pierre/IvorySQL/src/pl/plpgsql/src/expected/plpgsql_array.out 2026-01-23 10:15:32.156479475 +0100 +++ /home/pierre/IvorySQL/src/pl/plpgsql/src/results/plpgsql_array.out 2026-01-23 15:25:17.283892995 +0100 @@ -11,43 +11,57 @@ NOTICE: a = {1,2,4} do $$ declare a int[]; begin a := array[1,2]; a[3] := 4; raise notice 'a = %', a; end$$ language plisql; -NOTICE: a = {1,2,4} +ERROR: when ivorysql.compatible_mode is pg, not support a plisql procedural-language function statement +LINE 2: begin a := array[1,2]; a[3] := 4; raise notice 'a = %', a; e... + ^ do $$ declare a int[]; begin a[3] := 4; raise notice 'a = %', a; end$$ language plpgsql; NOTICE: a = [3:3]={4} do $$ declare a int[]; begin a[3] := 4; raise notice 'a = %', a; end$$ language plisql; -NOTICE: a = [3:3]={4} +ERROR: when ivorysql.compatible_mode is pg, not support a plisql procedural-language function statement +LINE 2: begin a[3] := 4; raise notice 'a = %', a; end$$ language pli... + ^ It looks like that make installcheck needs an already running PG instance. Is this correct ? In this case I understand this is due to the fact that I cannot create a PG instance running in Oracle mode: see https://github.com/IvorySQL/IvorySQL/issues/1171

Jan. 23, 2026

Error message wording

by Pierre Forstmann

Hello, In backend/parser/parser.c we have following code section: if(mode >RAW_PARSE_PLPGSQL_ASSIGN3) { /* Clean up (release memory) */ scanner_finish(yyscanner); ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), errmsg("when ivorysql.compatible_mode is pg, not support a plisql procedural-language function statement"))); } I think this error message should be rewritten in a better English like: when ivorysql.compatible_mode is set to pg, any plisql procedural-language function statement is not supported. Do you agree ? If yes, can I work on the related patch ?

Jan. 23, 2026