Form body encoding changed from + to %20 in 4.12.0 about okhttp

[swankjesse]

I don’t believe the behavior has changed in OkHttp 4.12.0.

We’ve got a test case in 4.12.0 that hasn’t changed since December 8, 2019.
https://github.com/square/okhttp/blob/okhttp_4x/okhttp/src/test/java/okhttp3/FormBodyTest.java#L26

But we did change it in 5.0.0-alpha.2, and I announced that in our changelog:
https://github.com/square/okhttp/pull/6518/files
https://square.github.io/okhttp/changelogs/changelog/#version-500-alpha2

Is it possible you’re running the 5.0 release unexpectedly, perhaps via a transitive dependency?

Either way I’m sorry for the disruption, and we don’t make breaking changes lightly.

[jimshowalter]

It appears we mixed mockwebserver alpha 5 with okhttp 4.12.0.

Oddly, it works, but we'd prefer to use com.squareup.okhttp3:mockwebserver:4.12.0.

But when we try to do that, there's no okhttp3.mockwebserver.MockWebServerExtension.

Is there some other non-5/non-alpha jar we need to add to get the equivalent of a JUnit 5 extension class for mockwebserver 4.12.0?

[jimshowalter]

We'd love to use your v5 alpha jars, but we have to keep dependencies aligned with the latest Spring and Spring Boot, and they're on 4.12.0.

[swankjesse]

One of the main goals of OkHttp 5.x is to fix our longstanding bug where MockWebServer is strongly coupled with JUnit 4.

[swankjesse]

I’ll do what I can to get 5.0 final out soon, which will hopefully be taken up quickly by our Spring friends.

[jimshowalter]

Is there some way to use mockwebserver alpha 5 without changing the form body?

[swankjesse]

Could you please elaborate on the problem with the new behaviour? According to the PR that changed it, the new behaviour is closer to what the specs expect.

[RonakWithCode]

In version 4.12.0, there was a change in form body encoding where spaces are now represented as %20 instead of the plus sign (+). Check the release notes or documentation for details, and update your code accordingly for compatibility.

[swankjesse]

@RonakWithCode there was no such behaviour change in 4.12. Please see the comment above.

[jimshowalter]

The mystery is that:

• We're on 4.12.0 for everything except mockwebserver.
• We have Gradle forces for okhttp, and they work even for mockwebserver, as shown in the output from gradlew dependencies:

--- com.squareup.okhttp3:mockwebserver3-junit5:5.0.0-alpha.2
+--- com.squareup.okhttp3:mockwebserver3:5.0.0-alpha.2
| +--- com.squareup.okhttp3:okhttp:5.0.0-alpha.2 -> 4.12.0

• The behavior of our form-body test changed from + to %20.
• Our form-body test doesn't even use mockwebserver:

package com.foo;

import jakarta.annotation.Nullable;
import java.io.File;
import java.io.IOException;
import java.nio.charset.Charset;
import okhttp3.MediaType;
import okhttp3.RequestBody;
import okhttp3.internal.Util;

public abstract class FooRequestBody {...

package com.foo;

import java.io.IOException;
import java.nio.charset.Charset;
import java.nio.charset.IllegalCharsetNameException;
import java.nio.charset.UnsupportedCharsetException;
import okhttp3.FormBody;

public class FooFormBody extends FooRequestBody {...

package com.foo;

import static org.junit.jupiter.api.Assertions.assertEquals;

import java.io.IOException;
import okio.Buffer;
import org.junit.jupiter.api.Test;

class FooFormBodyTest {
private FooFormBody body;

[yschimke]

Can you reproduce in a minimal build with only 4.12.0?

Otherwise it points to an issue with your gradle build still lying about the version.

[yschimke]

Closing for now as we have tests for this, will reopen if we can repro it failing on 4.12.0