fix(unstable): various resolution bugs for npm: specifiers

[bartlomieju]

1. cargo run -- run -A --unstable npm:vite build --debug

Deno has panicked. This is a bug in Deno. Please report this
at https://github.com/denoland/deno/issues/new.
If you can reliably reproduce this panic, include the
reproduction steps and re-run with the RUST_BACKTRACE=1 env
var set and include the backtrace in your report.

Platform: macos aarch64
Version: 1.24.3
Args: ["../deno/target/debug/deno", "run", "-A", "--unstable", "npm:vite", "build", "--debug"]

thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: could not find npm package 'source-map-support' referenced by '[email protected]'', ext/node/resolution.rs:595:6
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

2. cargo run -- run -A --unstable npm:vite@2 build --debug

error getting response  https://registry.npmjs.org/ Err(
    Error("missing field `name`", line: 1, column: 470),
)
error preparing Error {
    context: "Error getting response at https://registry.npmjs.org/",
    source: Error("missing field `name`", line: 1, column: 470),
}
failed to load config from /Users/ib/dev/my-vue-app/vite.config.js
error during build:
TypeError: Error getting response at https://registry.npmjs.org/
    at async loadConfigFromFile (file:https:///Users/ib/Library/Caches/deno/npm/registry.npmjs.org/vite/2.9.15/dist/node/chunks/dep-689425f3.js:61925:31)
    at async resolveConfig (file:https:///Users/ib/Library/Caches/deno/npm/registry.npmjs.org/vite/2.9.15/dist/node/chunks/dep-689425f3.js:61456:28)
    at async doBuild (file:https:///Users/ib/Library/Caches/deno/npm/registry.npmjs.org/vite/2.9.15/dist/node/chunks/dep-689425f3.js:41564:20)
    at async build (file:https:///Users/ib/Library/Caches/deno/npm/registry.npmjs.org/vite/2.9.15/dist/node/chunks/dep-689425f3.js:41552:16)
    at async CAC.<anonymous> (file:https:///Users/ib/Library/Caches/deno/npm/registry.npmjs.org/vite/2.9.15/dist/node/cli.js:738:9)

error preparing CustomError {
    class: "TypeError",
    message: "Relative import path \"@vitejs/plugin-vue\" not prefixed with / or ./ or ../\n    at file:https:///Users/ib/dev/my-vue-app/vite.config.js?t=1661250685827:2:17",
}
failed to load config from /Users/ib/dev/my-vue-app/vite.config.js
error during build:
TypeError: Relative import path "@vitejs/plugin-vue" not prefixed with / or ./ or ../
    at file:https:///Users/ib/dev/my-vue-app/vite.config.js?t=1661250685827:2:17
    at async loadConfigFromFile (file:https:///Users/ib/Library/Caches/deno/npm/registry.npmjs.org/vite/2.9.15/dist/node/chunks/dep-689425f3.js:61925:31)
    at async resolveConfig (file:https:///Users/ib/Library/Caches/deno/npm/registry.npmjs.org/vite/2.9.15/dist/node/chunks/dep-689425f3.js:61456:28)
    at async doBuild (file:https:///Users/ib/Library/Caches/deno/npm/registry.npmjs.org/vite/2.9.15/dist/node/chunks/dep-689425f3.js:41564:20)
    at async build (file:https:///Users/ib/Library/Caches/deno/npm/registry.npmjs.org/vite/2.9.15/dist/node/chunks/dep-689425f3.js:41552:16)
    at async CAC.<anonymous> (file:https:///Users/ib/Library/Caches/deno/npm/registry.npmjs.org/vite/2.9.15/dist/node/cli.js:738:9)

[bartlomieju]

There should be assertions verifying that name is not an empty string.

[bartlomieju]

cargo run -- run -A --unstable npm:prisma --help doesn't print anything. Adding console log into binary entry point is never printed which suggests that the code is never executed.

[bartlomieju]

No 1. is now addressed, but running it again yields another error:

error: Uncaught (in promise) TypeError: Loading unprepared module: https://deno.land/[email protected]/node/module.ts

[bartlomieju]

cargo run -- run -A --unstable npm:prisma --help doesn't print anything. Adding console log into binary entry point is never printed which suggests that the code is never executed.

This is now fixed as well

[bartlomieju]

This could use better error message instead of Relative import path "xo" is not prefix with /, ./ or ../

[dsherret]

I will open a PR with this change later.

[dsherret]

LGTM, but I think we should move referrer_package.dependencies.get to be first instead of last in that one method.

[dsherret]

I will open a PR with this change later.