fix(opencode): DB migrating on every run for non-latest channels

[bew]

Issue for this PR

Closes#21057
Closes#16885
Replaces #16884

Type of change

• Bug fix
• New feature
• Refactor / code improvement
• Documentation

What does this PR do?

This PR fixes DB migration that was running on every start of opencode when installed from nixpkgs.
The nix package in nixpkgs sets OPENCODE_CHANNEL = "stable", which makes the DB have the name opencode-stable.db instead of opencode.db that the marker implied.

👉 Using the actual database path which already takes the OPENCODE_CHANNEL into account fixes the problem.

(change done manually, zero AI involved)

How did you verify your code works?

Yes, locally.

Screenshots / recordings

The first run is from `main`, the second is with this PR.

Checklist

• I have tested my changes locally
• I have not included unrelated changes in this PR

[github-actions]

Thanks for your contribution!

This PR doesn't have a linked issue. All PRs must reference an existing issue.

Please:

1. Open an issue describing the bug/feature (if one doesn't exist)
2. Add Fixes #<number> or Closes #<number> to this PR description

See CONTRIBUTING.md for details.

[github-actions]

Thanks for updating your PR! It now meets our contributing guidelines. 👍

[bew]

The failing test / unit (linux) (pull_request) pipeline seems to be for another reason 🤔

1 tests failed:
(fail) memory: abort controller leak > webfetch does not leak memory over many invocations [487.59ms]

[jaredmontoya]

@bew what if the test is flaky? Can you rebase and force push so that CI runs again?

[bew]

@bew what if the test is flaky? Can you rebase and force push so that CI runs again?

I rebased on dev, let me know if there is anything else I can do!

[blake-hamm]

This branch is working for me on nixos with. TY!

[jaredmontoya]

@thdxr can this PR get a review and hopefully get merged please?
It seems like tests fail for reasons unrelated to the changes in this PR.

[jhol]

I ported this patch to 1.14.19 - the version I have via nixpkgs. It fixes the issue for me.

[bew]

I rebased on latest dev branch, which changed a few import paths.
Also added a proper commit description for this fix.

[bew]

Hello, I rebased on latest dev branch, any way to actually get this merged?

[jaredmontoya]

@simonklee@thdxr@adamdotdevin@Hona@Brendonovich@nexxeln anyone?
Don't let this simple PR rot here even longer than 40 days for no reason please.

[pwiens]

Heads up: this will silently break after rebase on fc34c7456 (#27615, May 15). That refactor renamed Database.Path (const) → Database.getPath() (function) and updated the four call sites in packages/opencode/src/cli/cmd/db.ts, but the marker in this PR isn't on main yet so it wasn't included.

Post-rebase, marker is undefined. fs.existsSync(undefined) returns false in both Bun 1.3.13 and Node 24, so !exists(undefined) is always true and the migration runs on every launch. The Bun build is transpile-only, so the missing namespace property doesn't surface at build time.

One-liner:

- if (!(await Filesystem.exists(Database.Path))) {+ if (!(await Filesystem.exists(Database.getPath()))) {

Cost is more than a noisy log line: JsonMigration.run re-scans the full JSON storage tree (6662 todos plus sessions/messages/parts in my case) on every launch. Not destructive thanks to onConflictDoNothing, but several seconds of disk IO at startup.

I hit this in a downstream fork that pulled in the same fix you have here, then rebased past fc34c7456.

[bew]

Thanks @pwiens, I updated the PR

[mmonfox]

I have the same issue, agent recommended `ln -s opencode-stable.db '~/.local/share/opencode/opencode.db' on my system and referenced this PR.

[bew]

Hello everyone!
In #30461 the json migration system was removed (first released in v1.16.0).
👉 So this PR is not needed anymore!