Tags: perki/backloop.dev-vite

2.3.0

2.3.0 as published to npm

Refresh the lockfile for 2.2.0

It could not resolve ^5.0.0 until backloop.dev@5.0.0 was on the registry, so it
was stuck claiming 2.0.0 and ^4.0.0 from before the split.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Claim the shared flag so backloop.dev does not repeat this

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Bump to 2.1.1 for the single-warning change

A Vite project loaded both packages and each said the same thing in turn. The
published build now claims a shared flag so backloop.dev does not repeat it;
sixteen lines become eight.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Claim the shared flag so backloop.dev does not repeat this

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Tag what is published, so the artifact has a git ref

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Say, on npm only, that npm is no longer where this is updated

This branch exists to be published. It is main plus this one file, and it must
stay that way.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Load the certificate in the hook, not at import

`vite build` failed outright, and had for a while:

  Error [ERR_REQUIRE_ASYNC_MODULE]: require() cannot be used on an ESM graph
  with top-level await.

Vite loads vite.config.js with require(). The config imports this plugin, the
plugin imported `backloop.dev`, and that package's ESM entry awaits the
certificate at the top level — so the config could not be loaded at all on
Node 22 or later. `apply: 'serve'` never helped: it gates the hooks, and by then
the import has already run.

Even where it did load, a production build downloaded a certificate it had no
use for, which since 4.0.0 also means it needed a secret to build.

Loading through the CommonJS entry inside the `config` hook fixes both: no
top-level await in the graph, and nothing fetched unless the dev server is
actually starting. Verified both ways — `vite build` completes with no request
made, `vite dev` serves HTTPS on *.backloop.dev with a valid handshake.

Not a regression from the 2.0.0 split: 1.0.12 against backloop.dev v3 fails
identically.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Refresh lockfile for v2.0.0

It still claimed version 1.0.12 and backloop.dev ^3.0.3 from the pre-split era,
and could not be regenerated until backloop.dev@4.0.0 was on the registry.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>