Invalid behaviour according to RFC.

[ioquatix]

sinatra/test/request_test.rb

Lines 20 to 27 in 4162502

	it 'raises Sinatra::BadRequest when multipart/form-data request has no content' do
	request = Sinatra::Request.new(
	'REQUEST_METHOD' => 'POST',
	'CONTENT_TYPE' => 'multipart/form-data; boundary=dummy',
	'rack.input' => StringIO.new('')
	)
	assert_raises(Sinatra::BadRequest) { request.params }
	end

This test is invalid according to https://www.rfc-editor.org/rfc/rfc2046#section-5.1.

I'm assuming this is because of golang/go#52519?

[dentarg]

It was added in #1743 to address #1739

As stated in that issue it was really easy to make any rack app blow up

I'm running my apps with the middleware at #1739 (comment) to avoid that happening

[dentarg]

Grape has a similar test I think? https://github.com/ruby-grape/grape/pull/2161/files#diff-ec432e9828fcad2ca5ef8aa6ac13454b84782708f4d2a5a8c23de9342d329d1a

[ioquatix]

@dentarg maybe you can give your opinion on rack/rack#1994 (comment) and the related issues.

[dentarg]

Sure

[dentarg]

Or well, my opinion is that we can not expect everyone on the internet to be following RFCs :-) We need to guard our apps against non-standard behaviour. I think that's what Sinatra is doing here? Sure, it would be better to address it directly in Rack.

[dentarg]

I think we have an understanding why this test case exist in Sinatra, can we close this issue @ioquatix?

It was needed after rack/rack#1572 that closed rack/rack#761.

[ioquatix]

You can close it, but if we adopt some mechanism to indicate "bad request" you may want to change the behaviour.

[ioquatix]

In Rack 3.1, this will exist: rack/rack#2019