[shane615]

Description

pg returns BIGINT (int8) as a string; SQLite and mysql2 return a number. All timestamps (created_at, closed_at, last_activity, …) are BIGINT, so on PostgreSQL they reach the code as text. generateTranscript() does new Date(ticketInfo.created_at), which gives an Invalid Date for a numeric string, so every close fails to build a transcript:

[performClose] Transcript generation error: RangeError: Invalid time value
    at formatDate (src/utils/transcript.js)

Nothing goes to the log channel and nothing is stored. The fix gives the PostgreSQL pool a type parser that returns int8 as a number (that pool only, pg defaults untouched). Every BIGINT column is a millisecond timestamp or a row id, far below Number.MAX_SAFE_INTEGER; Discord ids are VARCHAR, so there's no precision risk. NUMERIC results (AVG) keep the existing num() coercion.

Related Issue

None; found while testing #32–#34 live on PostgreSQL 16.

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)

How Was This Tested?

• Node.js 24.20.0, npm test: all pass, plus tests/postgres-int8.test.js (fails without the fix, needs no database server).
• Live on PostgreSQL 16: before the fix every close logged the error above; afterwards transcripts are posted to the log channel and stored. Checked read-only against the live database that ids, timestamps and COUNT(*) come back as numbers.

Checklist

• Code follows the project's conventions
• No new dependencies
• No new user-facing strings
• No schema change
• Works without MSK_API_KEY
• Verified on a test server
• CHANGELOG.md entry under [Unreleased]
• Conventional Commits

https://claude.ai/code/session_01HZPZxD3edzJNFFsiWsHYY1

[Musiker15]

Thanks, this one is clean too. The type parser only applies to our pool, and the counter-check works: the test fails without the fix.

#32 is merged now, so there's a conflict with main. I wanted to resolve that for you, but your fork belongs to an organization and GitHub doesn't let maintainers push to those. Could you rebase onto main? Two spots:

• CHANGELOG.md: keep your entry at the end of the ### Fixed list in [Unreleased].
• src/database/drivers/postgres.js: fix: robustness — DB outages, crashing events, shutdown and orphaned tickets #32 changed the Pool line, so types: typeParsers goes into that line:

pool = new Pool({ connectionTimeoutMillis: 10_000, ...descriptor.options, max: 10, types: typeParsers });

Keep the pool.on('error', ...) listener from #32 below it. Your test still passes with that, since the fake pool already has on(). I'll merge #38 first, so after that you may get one more small changelog conflict here.

[shane615]

Rebased onto main: types: typeParsers is in #32's Pool line with the pool.on('error', …) listener kept below it, and the changelog entry is at the end of ### Fixed. Tests pass. I'll rebase again if the changelog conflicts after #38.

[Musiker15]

Thanks for the quick rebase, the Pool line looks exactly right. #38 is merged now, so there's one more small conflict in CHANGELOG.md. Could you rebase once more and put your entry after the #38 one at the end of ### Fixed? I'll merge it right after.

[shane615]

Rebased again; the entry now follows the #38 one at the end of ### Fixed. Tests pass.