Customizer null checks in HTTP client builders assert the wrong attribute

[kdomo]

AbstractClientHttpRequestFactoryBuilder.mergedCustomizers(Consumer<T>) and its reactive counterpart assert on the customizers field instead of the customizer parameter:

protected final List<Consumer<T>> mergedCustomizers(Consumer<T> customizer) {
    Assert.notNull(this.customizers, "'customizer' must not be null");
    return merge(this.customizers, List.of(customizer));
}

The field can never be null because the constructor defaults it to an empty list, so the assertion always passes and a null customizer is not rejected. Callers get an NPE from List.of(customizer) on the next line rather than the intended IllegalArgumentException. The mergedCustomizers(Collection<Consumer<T>>) overload just below does check its parameter.

This affects withCustomizer(...) on all builders that extend these two base classes.

The same commit also corrects two Assert messages that name a parameter which doesn't exist on the method:

• ClientHttpConnectorBuilder.of(Class<T> clientHttpConnectorType) reports 'requestFactoryType', which looks like a leftover from ClientHttpRequestFactoryBuilder.of. The Assert.isTrue on the following line already uses the correct name.
• HttpComponentsClientHttpConnectorBuilder.withHttpClientCustomizer(Consumer<HttpAsyncClientBuilder> httpClientCustomizer) reports 'customizer', while withConnectionManagerCustomizer and withConnectionConfigCustomizer in the same class both use their parameter names.

I've added tests for the null checks. There's no test class for the ClientHttpConnectorBuilder interface, so the of message change isn't covered — happy to add one if you'd prefer.

[snicoll]

Thanks. The tests should have been put on the base class for consistency. In doing so I've noticed one more place where that check isn't consistent. I have also removed the update of assertion in the two other locations that are unrelated.

See 60445a5