`PGPORT` environment variable doesn't seem to be respected in v1.3.0

[cfeenstra67]

Hi there guys, wanted to alert you to an issue that I started facing starting after the v1.3.0 release on Friday. I use sqitch via the docker image & the recommended script as documented in the README and docker hub. I am not sure if this an issue with the v1.3.0 release in general or just the docker setup. I'm also not sure if it's all PG* environment variables or just PGPORT. Basically, the behavior I'm seeing is that even though PGPORT is in the environment, when trying to run migrations I get an error like this:

Adding registry tables to pg-localdev
psql: error: could not connect to server: Connection refused
        Is the server running on host "localhost" (::1) and accepting
        TCP/IP connections on port 5432?
could not connect to server: Connection refused
        Is the server running on host "localhost" (127.0.0.1) and accepting
        TCP/IP connections on port 5432?
"psql" unexpectedly returned exit value 2

Steps to reproduce:

1. Have a postgres instance running on port 5433
2. Download the sqitch docker running script from the README of this repo
3. Try to run some migrations with PGPORT=5433 in the environment

This is what my sqitch config file looks like:

[core]
  engine = pg
  top_dir = db/
[engine "pg"]
  target = pg-localdev
[target "pg-localdev"]
  uri = db:postgresql://postgres:password@localhost/database

I was able to solve this issue for myself by just pinning the tag in the sqitch runner script to v1.2.1.1 (side note: v1.2.1 is listed as a supported tag on docker hub but doesn't seem to work: https://hub.docker.com/r/sqitch/sqitch/ ; I had to come look at the tags in the repo to find that you use a versioning scheme w/ 4 components instead of 3). So I'm not blocked here or anything, just thought you guys might want a heads up as this caused some github actions to start failing for me; I am using their "port mapping" functionality (https://docs.github.com/en/actions/using-containerized-services/about-service-containers) and was specifying the port with PGPORT, and my tests suddenly started failing. I'm guessing the same might be true for some others as well.

[theory]

Confirmed this is an issue:

$ PGPORT=5433 PGUSER=david ~/dev/sqitch/docker-sqitch/docker-sqitch.sh deploy db:pg://host.docker.internal/porttest
Adding registry tables to db:pg://host.docker.internal/porttest
psql: error: could not connect to server: Connection refused
	Is the server running on host "host.docker.internal" (192.168.65.2) and accepting
	TCP/IP connections on port 5432?
"psql" unexpectedly returned exit value 2

Workaround is to include the port in the target, not just the environment variable:

PGPORT=5433 PGUSER=david ~/dev/sqitch/docker-sqitch/docker-sqitch.sh deploy db:pg://host.docker.internal:5433/porttest
Adding registry tables to db:pg://host.docker.internal:5433/porttest
Deploying changes to db:pg://host.docker.internal:5433/porttest
  + change1 .. ok
  + change2 .. ok

Does not seem to apply to PGDATABASE, which properly replaces the database name when it's missing from the target URL:

$ PGPORT=5433 PGUSER=david PGDATABASE=porttest ~/dev/sqitch/docker-sqitch/docker-sqitch.sh deploy db:pg://host.docker.internal:5433/
Deploying changes to db:pg://host.docker.internal:5433/porttest
  + change1 .. ok
  + change2 .. ok

So it seems specific to PGPORT.

[theory]

Figured it out, broken in 160b54ef0. Will switch back to calling _port in the next version.