Add PGXS "make installcheck" support

[df7cb]

This enables running PG's standard testsuite with a PG server configured
for decoderbufs.

Example on Debian:

$ sudo apt install postgresql-14-decoderbufs
$ pg_virtualenv -o shared_preload_libraries=decoderbufs -o wal_level=logical
Creating new PostgreSQL cluster 14/regress ...
$ make installcheck
/usr/lib/postgresql/14/lib/pgxs/src/makefiles/../../src/test/regress/pg_regress --inputdir=./ --bindir='/usr/lib/postgresql/14/bin'    --dbname=contrib_regression decoderbufs
(using postmaster on localhost, port 5433)
============== dropping database "contrib_regression" ==============
HINWEIS:  Datenbank »contrib_regression« existiert nicht, wird übersprungen
DROP DATABASE
============== creating database "contrib_regression" ==============
CREATE DATABASE
ALTER DATABASE
============== running regression test queries        ==============
test decoderbufs                  ... ok           30 ms

=====================
 All 1 tests passed.
=====================

$ exit
Dropping cluster 14/regress ...

Tested on apt.postgresql.org with PG 10+ on amd64/arm64/i386/ppc64el.

https://pgdgbuild.dus.dg-i.net/view/Testsuite/job/postgres-decoderbufs-autopkgtest/

[df7cb]

Is anyone listening here? PG 17 will be released on Thursday, and decoderbufs isn't ready yet:

make[1]: Verzeichnis „/home/myon/projects/postgresql/postgres-decoderbufs/postgres-decoderbufs“ wird betreten
gcc -Wall -Wmissing-prototypes -Wpointer-arith -Wdeclaration-after-statement -Werror=vla -Wendif-labels -Wmissing-format-attribute -Wimplicit-fallthrough=3 -Wcast-function-type -Wshadow=compatible-local -Wformat-security -fno-strict-aliasing -fwrapv -fexcess-precision=standard -Wno-format-truncation -Wno-stringop-truncation -g -g -O2 -Werror=implicit-function-declaration -fstack-protector-strong -fstack-clash-protection -Wformat -Werror=format-security -fcf-protection -fno-omit-frame-pointer -g -O2 -Werror=implicit-function-declaration -ffile-prefix-map=/home/myon/projects/postgresql/postgres-decoderbufs/postgres-decoderbufs=. -fstack-protector-strong -fstack-clash-protection -Wformat -Werror=format-security -fcf-protection -fPIC -fvisibility=hidden -std=c11  -I/usr/local/include  -I. -I./ -I/usr/include/postgresql/17/server -I/usr/include/postgresql/internal  -Wdate-time -D_FORTIFY_SOURCE=2 -D_GNU_SOURCE -I/usr/include/libxml2   -c -o src/decoderbufs.o src/decoderbufs.c
src/decoderbufs.c: In function ‘pg_decode_change’:
src/decoderbufs.c:625:53: error: ‘HeapTupleData’ has no member named ‘tuple’
  625 |                            &change->data.tp.newtuple->tuple, tupdesc);
      |                                                     ^~
src/decoderbufs.c:631:53: error: ‘HeapTupleData’ has no member named ‘tuple’
  631 |                            &change->data.tp.newtuple->tuple, tupdesc);
      |                                                     ^~
src/decoderbufs.c:646:55: error: ‘HeapTupleData’ has no member named ‘tuple’
  646 |                              &change->data.tp.oldtuple->tuple, tupdesc);
      |                                                       ^~
src/decoderbufs.c:656:55: error: ‘HeapTupleData’ has no member named ‘tuple’
  656 |                              &change->data.tp.newtuple->tuple, tupdesc);
      |                                                       ^~
src/decoderbufs.c:662:55: error: ‘HeapTupleData’ has no member named ‘tuple’
  662 |                              &change->data.tp.newtuple->tuple, tupdesc);
      |                                                       ^~
src/decoderbufs.c:678:53: error: ‘HeapTupleData’ has no member named ‘tuple’
  678 |                            &change->data.tp.oldtuple->tuple, tupdesc);
      |                                                     ^~
make[1]: *** [<eingebaut>: src/decoderbufs.o] Fehler 1

[gunnarmorling]

Paging @jpechane :)

Most attention is on pgoutput these days, though, and it's the recommended decoding plug-in when available.

[df7cb]

Is that the pgoutput.so shipped with contrib? That's available on every Debian install. If that is strictly better, we could also drop the Debian package of decoderbufs and go with pgoutput instead.

[jpechane]

@df7cb Hi, we definitely want to keep protobuf for some time. We'll make pgoutput default option sometimes during 3.x timeline and deprecate protbuf later depending on community feedback.

[jpechane]

@df7cb Please see https://issues.redhat.com/browse/DBZ-8275

[df7cb]

3.0.1 doesn't compile here against PG17:

src/decoderbufs.c: In function ‘pg_decode_change’:
src/decoderbufs.c:631:28: error: passing argument 3 of ‘tuple_to_tuple_msg’ from incompatible pointer type [-Wincompatible-pointer-types]
  631 |                            &change->data.tp.newtuple TUPLE_ACCESS, tupdesc);
      |                            ^~~~~~~~~~~~~~~~~~~~~~~~~
      |                            |
      |                            HeapTupleData **
src/decoderbufs.c:396:61: note: expected ‘HeapTuple’ {aka ‘HeapTupleData *’} but argument is of type ‘HeapTupleData **’
  396 |                                Relation relation, HeapTuple tuple,
      |                                                   ~~~~~~~~~~^~~~~

Also, could you perhaps enable GitHub issues on this repository? It's the only one I know that has only pull requests enabled, but not issues.

[jpechane]

@df7cb We do use Jira exclusively. It would be a bit of complications for us to track issue accross mutliple issue trackers.

[bradfordboyle]

@df7cb PR #32 should address the compilation error with PG17. I've tested this change on trixie with both PG16 and PG17 using the installcheck test that you've added in this PR.

[jpechane]

@df7cb Hi, I am sorry I am necormancing this PR. I'd like to get it in but if possible also add a github workflow that would run on each PR and push to main. The workflow would build the plguin and run the installcheck on it. Would it be possible?

[df7cb]

@df7cb PR #32 should address the compilation error with PG17. I've tested this change on trixie with both PG16 and PG17 using the installcheck test that you've added in this PR.

Thanks, the Debian packages are happy now.

[df7cb]

@df7cb Hi, I am sorry I am necormancing this PR. I'd like to get it in but if possible also add a github workflow that would run on each PR and push to main. The workflow would build the plguin and run the installcheck on it. Would it be possible?

There's various extensions doing that, for example https://github.com/df7cb/postgresql-unit/blob/master/.github/workflows/regression.yml

[Naros]

❌ Developer Certificate of Origin (DCO) check failed.

Hi @df7cb, please sign off all commits with:

git commit -s

If pull request commits are not signed off, the pull request cannot be merged. For more information about why this is required, please see our blog about contribution requirement changes.

[df7cb]

@Naros that blog post is from 2025 while this PR is from 2022. I've since changed employers and this PR is coming from their namespace where I no longer have access to. Please just accept it anyway. You can copy the DCO from here:

Signed-off-by: Christoph Berg <[email protected]>

Thanks.

[Naros]

Thanks @df7cb. I'll defer to @jpechane here, but I think your comment suffices for our needs.