{testsdk} Add ContentLengthProcessor to fix Content-Length header

[jiasli]

Description

When the test framework replaces a random name, its length is not preserved:

- appxxxxxxxxxxxx
+ app000003

This make Content-Length header in the response not match the actual body, causing failure due to the new azure-core check Azure/azure-sdk-for-python#20888.

There are several processors that changes the Content-Length of the body:

1. All preparers that inherits from AbstractPreparer or SingleValueReplacer, like ResourceGroupPreparer
2. AADGraphUserReplacer
3. GraphClientPasswordReplacer
4. etc

This makes it difficult to enforce all processors to rectify the Content-Length header. The best place is to rectify Content-Length header when all processors are finished.

See

• {CI} Fix Content-Length header's value in recordings #20541
• {CI} Make recording responses' Content-Length match the body #20533

[jiasli]

We can't add ContentLengthProcessor here

azure-cli/src/azure-cli-testsdk/azure/cli/testsdk/base.py

Lines 89 to 97 in 30216d0

	default_recording_processors = [
	SubscriptionRecordingProcessor(MOCKED_SUBSCRIPTION_ID),
	AADAuthRequestFilter(),
	LargeRequestBodyProcessor(),
	LargeResponseBodyProcessor(),
	DeploymentNameReplacer(),
	RequestUrlNormalizer(),
	self.name_replacer
	] + self._processors_to_reset

because AbstractPreparer will add preparers when executing the test method

azure-cli/src/azure-cli-testsdk/azure/cli/testsdk/scenario_tests/preparers.py

Line 34 in 30216d0

test_class_instance.recording_processors.append(self)

causing these newly added preparers to be called after ContentLengthProcessor

[wangzelin007]

Why response['headers']['content-length'] is list while request.headers['Content-Length'] is str ?

[jiasli]

Very good question. Because request is a vcr.request.Request object while response is a dict. vcrpy defines different data structures for them, though I don't know why.