Re: In-placre persistance change of a relation

Hi,

On Fri, Jan 14, 2022 at 11:43:10AM +0900, Kyotaro Horiguchi wrote:
> I found a bug.
> 
> mdmarkexists() didn't close the tentatively opend fd. This is a silent
> leak on Linux and similars and it causes delete failure on Windows.
> It was the reason of the CI failure.
> 
> 027_persistence_change.pl uses interactive_psql() that doesn't work on
> the Windos VM on the CI.
> 
> In this version the following changes have been made in 0001.
> 
> - Properly close file descriptor in mdmarkexists.
> 
> - Skip some tests when IO::Pty is not available.
>   It might be better to separate that part.
> 
> Looking again the ALTER TABLE ALL IN TABLESPACE SET LOGGED patch, I
> noticed that it doesn't implement OWNED BY part and doesn't have test
> and documenttaion (it was PoC). Added all of them to 0002.

The cfbot is failing on all OS with this version of the patch.  Apparently
v16-0002 introduces some usage of "testtablespace" client-side variable that's
never defined, e.g.
https://api.cirrus-ci.com/v1/artifact/task/4670105480069120/regress_diffs/src/bin/pg_upgrade/tmp_check/regress/regression.diffs:

diff -U3 /tmp/cirrus-ci-build/src/test/regress/expected/tablespace.out /tmp/cirrus-ci-build/src/bin/pg_upgrade/tmp_check/regress/results/tablespace.out
--- /tmp/cirrus-ci-build/src/test/regress/expected/tablespace.out	2022-01-18 04:26:38.744707547 +0000
+++ /tmp/cirrus-ci-build/src/bin/pg_upgrade/tmp_check/regress/results/tablespace.out	2022-01-18 04:30:37.557078083 +0000
@@ -948,76 +948,71 @@
 CREATE SCHEMA testschema;
 GRANT CREATE ON SCHEMA testschema TO regress_tablespace_user1;
 CREATE TABLESPACE regress_tablespace LOCATION :'testtablespace';
+ERROR:  syntax error at or near ":"
+LINE 1: CREATE TABLESPACE regress_tablespace LOCATION :'testtablespa...