Package doesn't work with --disable-proto=delete

[dsherret]

When I run the unit tests with NODE_OPTIONS=--disable-proto=delete I get the following error:

$ npm run tests-only

> [email protected] tests-only
> NODE_OPTIONS=--disable-proto=delete nyc tape 'test/**/*.js'

/mnt/v/typed-array-byte-offset/index.js:6
++cov_1i9ps04zcs.s[16];descriptor=gOPD(superProto,'byteOffset');}else{++cov_1i9ps04zcs.b[3][1];}// Opera 12.16 has a magic byteOffset data property on instances AND on Proto
                                  ^

TypeError: Cannot convert undefined or null to object
    at getOwnPropertyDescriptor (<anonymous>)
    at /mnt/v/typed-array-byte-offset/index.js:6:35
    at forEachArray (/mnt/v/typed-array-byte-offset/node_modules/for-each/index.js:12:17)
    at forEach (/mnt/v/typed-array-byte-offset/node_modules/for-each/index.js:54:9)
    at Object.<anonymous> (/mnt/v/typed-array-byte-offset/index.js:1:7537)
    at Module._compile (node:internal/modules/cjs/loader:1369:14)
    at Module.replacementCompile (/mnt/v/typed-array-byte-offset/node_modules/nyc/node_modules/append-transform/index.js:58:13)
    at module.exports (/mnt/v/typed-array-byte-offset/node_modules/nyc/node_modules/default-require-extensions/js.js:8:9)
    at Object.<anonymous> (/mnt/v/typed-array-byte-offset/node_modules/nyc/node_modules/append-transform/index.js:62:4)
    at Module.load (node:internal/modules/cjs/loader:1206:32)

Node.js v20.12.2

I think this is because the has-proto only checks if __proto__ is allowed on object creation and not after the fact?

[ljharb]

You're right; the semantics of has-proto are definitely "can be set in an object literal" and not "the accessor exists".

I'm not sure I care to support this particular node flag - it's just not reasonable to enable it with how many parts of the ecosystem it breaks. It wouldn't be hard to add, but it'd add an extra dependency on reflect.getprototypeof.

[dsherret]

It wouldn't be hard to add, but it'd add an extra dependency

Would you be willing to support this scenario for where the Reflect global exists so a dependency doesn't need to be added (if __proto__ isn't supported then likely Reflect will exist)? Also, it seems the dependency on has-proto could be removed since it's not checking the right thing?

[ljharb]

Indeed, it'd remove the has-proto dep. My robustness constraints don't allow me to solely rely on Reflect, though - I'll probably end up adding the dep.

[dsherret]

Thanks! Not relying on __proto__ slowly moves the ecosystem in a better place.

[ljharb]

I think the accessor is fine as-is, personally, and removing it isn't "better", but correctness does matter, and it was clearly incorrect. (also, an environment without it isn't actually an environment built on web standards, since the web requires it to always be present)

v1.0.3 will be published later today.

[ljharb]

v1.0.3 is released. I'll also add some regression tests.