sql,pgwire: fix a host of pgwire/sql interaction bugs

[knz]

First commit from #30691.

Fixes #30701.
Fixes #30692.
Fixes #30693.
Fixes #30697.
Fixes #30698.
Fixes #30703.
Informs #30717.

With special thanks to @benesch for asking me to poke at this.

[cockroach-teamcity]

This change is 

[benesch]

that review was challenging—there's a lot crammed into the last commit—but I so appreciate you doing this. Note however that I didn't review the first commit, since Andrei appears to be already reviewing that in #30691.

Reviewed 3 of 3 files at r1, 1 of 1 files at r2, 4 of 4 files at r3, 29 of 29 files at r4.
Reviewable status: complete! 1 of 0 LGTMs obtained

---

pkg/sql/internal.go, line 265 at r4 (raw file):

	opName string,
	txn *client.Txn,
	sargs *SessionArgs,

Seems like this should maybe just take the username as a string argument? Perhaps I'm missing something.

---

pkg/sql/logictest/logic.go, line 918 at r4 (raw file):

	if _, err := db.Exec("SET extra_float_digits = 0"); err != nil {
		t.Fatal(err)
	}

🤦‍♂️ this can't have been fun to track down

---

pkg/sql/pgwire/conn.go, line 271 at r4 (raw file):

				return err
			}
		}

Is this enough of a performance win to be worth it? I don't love the idea of having a slightly-differently-configured client in unit tests than in production.

[knz]

Reviewable status: complete! 1 of 0 LGTMs obtained

---

pkg/sql/internal.go, line 265 at r4 (raw file):

Previously, benesch (Nikhil Benesch) wrote…

Seems like this should maybe just take the username as a string argument? Perhaps I'm missing something.

Good suggestion. Done.

---

pkg/sql/logictest/logic.go, line 918 at r4 (raw file):

Previously, benesch (Nikhil Benesch) wrote…

🤦‍♂️ this can't have been fun to track down

Of the two this was the simplest: we print out the list of session vars, their current value and their defautls in the "pg_catalog" logic test, and this was showing up as "2" where I was expecting "0". I then verified that "psql" leaves the default at 0 where "cockroach sql" turned it into "2" so I knew the problem was lib/pq.

The "not fun" one was the timestamp stuff -- that Go will format the timezone differently depending on whether it is "named" or not was a learning experience, to say the least.

---

pkg/sql/pgwire/conn.go, line 271 at r4 (raw file):

Previously, benesch (Nikhil Benesch) wrote…

Is this enough of a performance win to be worth it? I don't love the idea of having a slightly-differently-configured client in unit tests than in production.

This is because the test here needs to "plug in" its own query handler to intercept queries in lieu of the conn executor.

[benesch]

Reviewed 32 of 32 files at r5, 1 of 1 files at r6, 2 of 4 files at r7, 29 of 29 files at r8.
Reviewable status: complete! 1 of 0 LGTMs obtained

[knz]

bors r+

[craig]

Build succeeded

• GitHub CI (Cockroach)