Closed olirice closed 2 months ago
pcnc
commented
2 months ago Looks good - some tests seem to already be failing (intended?)
We could have the workflow output the regression output file as an artifact in case it exists, to surface the impacted changes more easily - this workflow points the regression test output as being located at /nix/store/q4bw6l6wkiz1hljvqwl5f93j7ifx1nnj-postgres-15.6-check-harness/regression_output/regression.out
steve-chavez
commented
2 months ago Getting a failure locally, some differences on plcoffee, plls, plv8. This is the diff:
diff -U3 /nix/store/7dmppm4f3a21iddmvy924llqxh60ipl0-tests/expected/extensions_sql_interface.out /nix/store/lnwy4z38namdhv859s2pc0jfnklgb42x-postgres-15.6-check-harness/regression_output/results/extensions_sql_interface.out
--- /nix/store/7dmppm4f3a21iddmvy924llqxh60ipl0-tests/expected/extensions_sql_interface.out 1970-01-01 00:00:01.000000000 +0000
+++ /nix/store/lnwy4z38namdhv859s2pc0jfnklgb42x-postgres-15.6-check-harness/regression_output/results/extensions_sql_interface.out 2024-08-06 16:01:49.193370738 +0000
@@ -20,13 +20,10 @@
installed_version is null
order by
name asc;
- name
-----------
+ name
+---------
pg_cron
- plcoffee
- plls
- plv8
-(4 rows)
+(1 row)
/*
@@ -97,8 +94,11 @@
pgsodium | f
pgstattuple | t
pgtap | t
+ plcoffee | f
+ plls | f
plpgsql | f
plpgsql_check | f
+ plv8 | f
postgis | f
postgis_raster | f
postgis_sfcgal | t
@@ -120,7 +120,7 @@
vector | t
wrappers | f
xml2 | f
-(77 rows)
+(80 rows)
/*
@@ -3028,6 +3028,12 @@
pgtap | public | volatility_is | name, name[], text, text | text
pgtap | public | volatility_is | name, text | text
pgtap | public | volatility_is | name, text, text | text
+ plcoffee | pg_catalog | plcoffee_call_handler | | language_handler
+ plcoffee | pg_catalog | plcoffee_call_validator | oid | void
+ plcoffee | pg_catalog | plcoffee_inline_handler | internal | void
+ plls | pg_catalog | plls_call_handler | | language_handler
+ plls | pg_catalog | plls_call_validator | oid | void
+ plls | pg_catalog | plls_inline_handler | internal | void
plpgsql | pg_catalog | plpgsql_call_handler | | language_handler
plpgsql | pg_catalog | plpgsql_inline_handler | internal | void
plpgsql | pg_catalog | plpgsql_validator | oid | void
@@ -3053,6 +3059,12 @@
plpgsql_check | public | plpgsql_profiler_reset_all | | void
plpgsql_check | public | plpgsql_show_dependency_tb | fnname text, relid regclass | TABLE(type text, oid oid, schema text, name text, params text)
plpgsql_check | public | plpgsql_show_dependency_tb | funcoid regprocedure, relid regclass | TABLE(type text, oid oid, schema text, name text, params text)
+ plv8 | pg_catalog | plv8_call_handler | | language_handler
+ plv8 | pg_catalog | plv8_call_validator | oid | void
+ plv8 | pg_catalog | plv8_info | | json
+ plv8 | pg_catalog | plv8_inline_handler | internal | void
+ plv8 | pg_catalog | plv8_reset | | void
+ plv8 | pg_catalog | plv8_version | | text
postgis | public | _postgis_deprecate | oldname text, newname text, version text | void
postgis | public | _postgis_index_extent | tbl regclass, col text | box2d
postgis | public | _postgis_join_selectivity | regclass, text, regclass, text, text | double precision
@@ -5090,7 +5102,7 @@
xml2 | public | xpath_table | text, text, text, text, text | SETOF record
xml2 | public | xslt_process | text, text | text
xml2 | public | xslt_process | text, text, text | text
-(4940 rows)
+(4952 rows)
/*
olirice
commented
2 months ago Initial output was produced on an M1 Mac and CI is running ubuntu. I'm firing up an old linux box at home to reproduce the diff but that's probably the culprit
olirice
commented
2 months ago I commented out the procedural languages to unblock this PR. Those were the ones causing issues on Mac Theres an issue for it here https://github.com/supabase/postgres/issues/1094
Creates a test
/nix/test/sql/extensions_sql_interface.outthat monitors extension owned functions, tables, views, matviews, and indexes for changes as versions get bumped
We will use this to proactively monitor for breaking changes that could impact pause/restore before extensions go to production
Currently the test is very broad and includes functions etc that are prefixed with
_<something>implying that they are internal. If we see a lot of low impact churn in those functions we can exclude them but for now I figured it makes sense to be as cautious as possible.