TWO CHANGES, together because they touch the same packager hunks.
1. THE DRIVER.
The BrightSign package used to be MANUFACTURED. scripts/build-server-zip.sh dropped
better-sqlite3 from package.json and then installed db/sqlite-compat.js into
node_modules under that name, so every require resolved to the façade. It worked —
and it shipped a database layer that no test had ever executed. That is the same
shape as the TELEMETRY_COLLECTOR TDZ crash that took production down while 1676
tests and four CI jobs were green: a build-time rewrite cannot be tested by the
build that performs it.
db/sqlite-driver.js now decides at runtime: the native driver when it loads, the
node:sqlite façade otherwise. One artifact, one code path, and — the point — both
branches reachable from a test. ST_SQLITE_DRIVER=node runs the entire suite the way
a player runs it, and a new CI job does exactly that on Node 24 with --omit=optional
so the fallback is reached the same way it is on hardware, not by an env var alone.
better-sqlite3 becomes an optionalDependency, so a host with no compiler installs
cleanly and falls back rather than failing. preflight-deps stops trying to rebuild a
native module on a host that has no toolchain and a working built-in driver — on a
player that was a five-minute node-gyp failure ending in a server that never started.
Asking for the native driver BY NAME (ST_SQLITE_DRIVER=better-sqlite3) still fails
loudly, because a production box that has lost its native module is broken and should
say so rather than quietly running something else.
⚠️ NODE 24 IN PRACTICE. node:sqlite is unflagged only from 23.4; on the 22.x line it
needs --experimental-sqlite and on 20.x it does not exist. So the code probes rather
than comparing versions, the player package pins engines >=24, and the built-in cases
skip on the Node 20 CI job rather than failing there.
Verified on Node 24, both drivers, full suite:
better-sqlite3 1762 pass / 0 fail
node:sqlite 1762 pass / 0 fail
and the built payload resolves node:sqlite with no better-sqlite3 present at all.
2. THE LICENCE.
The ffprobe/ffmpeg binaries added in the previous commit are LGPL 2.1 and statically
linked, so the licence text has to travel WITH them — a link on a website is not the
copy the licence asks to accompany the work. The packager now copies
COPYING.LGPLv2.1 and a build README into bin/, and refuses to build if the licence is
missing. legal/third-party.html gains an LGPL section with the written offer required
by section 6 for static linking, and the exact configure line.
It also drops Sharp, which that page still listed although #263 removed it, and names
what actually does the image work now (jimp, @jsquash/webp, @jsquash/avif).
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014kfhrUPit5MCqxeTQyqr56
scripts/upgrade.sh already runs `npm ci`, so this is not for the normal path. It is
for the ways a box ends up with the wrong node_modules, both of which present as
"server will not start" with an error naming a file rather than the action needed:
ROLLBACK checking out an older tag to back out a bad release restores that
tag's package.json but not its packages. This branch removes
google-auth-library, so a rollback to main would not boot — and you
are rolling back because something else already broke.
NODE UPGRADE better-sqlite3 is compiled against one ABI. Upgrading Node makes every
boot fail with NODE_MODULE_VERSION, which reads like database
corruption and is not.
Runs as the FIRST statement in server.js, before any dependency is required, and uses
only Node builtins — anything it imported could be the thing that is missing. Repairs
with `npm install --omit=dev` (never `ci` on a partly-populated tree, which would
delete a working node_modules to fix one package) or `npm rebuild better-sqlite3`, and
exits with the command to run if it cannot. ST_SKIP_DEP_PREFLIGHT=1 opts out.
⚠️ The first version of the native check was WRONG and I caught it only by running it
under a real version mismatch: better-sqlite3's entry point is plain JavaScript that
loads the compiled binding lazily, so `require()` succeeds under a Node the binary was
never built for. It reported a genuinely broken install as healthy. It now opens an
in-memory database, which is what actually pulls the binding in. A test pins that,
because the failure is invisible — the check keeps passing on every machine where
nothing is wrong.
Verified: a deleted dependency is detected, installed and the server boots (200); the
ABI mismatch is detected under Node 18 against a module built for Node 20 and reported
clean under Node 20; a healthy tree costs 8ms and touches no network.
1609 tests pass.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Bvjey4FNam49MN7ybjcq6A
