Not getting the full type when parsing multipart/form-data

[jimmywarting]

Bug Description

Not getting the full type when parsing multipart/form-data

Reproducible By

var b = new Blob(['123'], { type: 'text/plain;charset=utf-8'})
var fd = new FormData()
fd.set('x', b)
var res = new Response(fd)
res.clone().text().then(body => 
  // Just making sure that it contains ;charset=utf-8 (which it dose)
  console.log(body.includes('utf-8'))
)

new Response(fd).formData().then(fd => {
  // returns just 'text/plain'
  console.log(fd.get('x').type)
})

Expected Behavior

To get the full type, text/plain;charset=utf-8

Logs & Screenshots

Screenshot of chrome:

Screenshot of NodeJS

Environment

Not using undici directly - using Node.js v20.6.1 built in fetch

Additional context

presumable this revolves around the usage around Busboy...
Maybe the type should be extracted from the headers instead...? to get the raw content-type.

[KhafraDev]

would you like to send in a PR?

[jimmywarting]

would you like to send in a PR?

Not this time... this time i think i will pass.
Have plenty of other things to do.

[barronw]

Hi @jimmywarting and @KhafraDev. Can I work on this issue?

[jimmywarting]

fine by me

[barronw]

The fix seems to require a change to busboy, and there is a pull request (#2211) to replace busboy with @fastify/busboy. I'll wait to see if that lands before submitting a fix.

[IlyasShabi]

@barronw Indeed, it comes from busboy, and even @fastify/busboy returns text/plain as the mimeType

[KhafraDev]

@fastify/busboy is a fork of busboy, the options are probably the same

[IlyasShabi]

@KhafraDev @barronw I think that we can emit charset also boy.emit('file', fieldname, file, filename, encoding, contype, charset) https://github.com/fastify/busboy/blob/master/lib/types/multipart.js#L193 and then handle the logic from in undici like:
busboy.on('file', (name, value, filename, encoding, mimeType, charset) https://github.com/nodejs/undici/pull/2211/files#diff-bd6db9a8bceb4e9fce3f93ba5284c81f46a44a1187909e8609006e1efeab2570R399

[jimmywarting]

@IlyasShabi
charset should not be the only thing we should worry about...

if i would like then i could just send this

var type = 'multipart/form-data; boundary=----WebKitFormBoundaryv9GiqHf1i48Tx1aI'
var otherFormData = new Blob(['123'], { type })
console.assert(otherFormData.type === type)
var fd = new FormData()
fd.set('embeded-formdata', otherFormData)

// later
fd = await request.formData()
console.assert(fd.get('embeded-formdata').type === type)

i should be able to put in a file that have completely nonsense type

new Blob(data, { type: 'x/y; foo=bar; fuz=123' })

and get back exactly the same thing.