FetchBackend expects case sensitive 'Content-Type' about angular

[diesieben07]

The XhrBackend behaves differently. It uses HttpHeaders#has, which is a case-insensitive check. Note also that this issue not only applies to Content-Type but also Accept.

See here in XhrBackend:

angular/packages/common/http/src/xhr.ts

Lines 97 to 109 in bac5ba3

	// Add an Accept header if one isn't present already.
	if (!req.headers.has('Accept')) {
	xhr.setRequestHeader('Accept', 'application/json, text/plain, */*');
	}
	// Auto-detect the Content-Type header if one isn't present already.
	if (!req.headers.has('Content-Type')) {
	const detectedType = req.detectContentTypeHeader();
	// Sometimes Content-Type detection fails.
	if (detectedType !== null) {
	xhr.setRequestHeader('Content-Type', detectedType);
	}
	}

See here for how HttpHeaders#has is case-insensitive:

angular/packages/common/http/src/headers.ts

Lines 98 to 99 in 0460a9d

	return this.headers.has(name.toLowerCase());
	}

See here for the problematic code in FetchBackend:

angular/packages/common/http/src/fetch.ts

Lines 247 to 257 in bac5ba3

	// Add an Accept header if one isn't present already.
	headers['Accept'] ??= 'application/json, text/plain, */*';
	// Auto-detect the Content-Type header if one isn't present already.
	if (!headers['Content-Type']) {
	const detectedType = req.detectContentTypeHeader();
	// Sometimes Content-Type detection fails.
	if (detectedType !== null) {
	headers['Content-Type'] = detectedType;
	}
	}

[diesieben07]

@wduminy The reason that binary data doesn't exhibit the same behavior is that Angular doesn't have a content type to infer.

[JeanMeche]

It looks like the behavior is consistent with how the XhrBackend is behaving. Although I agree headers should be case insensitive.

I'm wondering how breaking would it be to make that change.

[JeanMeche]

Right, I must have missed something in my tests.

#56539 will provide the fix for this. Thank you for the investigation.

[wduminy]

@JeanMeche Quick update: In the code I reported above I used JSON data; but when I used binary data -- a more realistic use case -- the problem did not occur -- i.e. application/json was not inserted.

So case sensitivity might be a problem; but the use-case here is certainly "contrived by accident".