The test verifies that saved passwords always are sent to be autofilled if they have the "generated" bit set. However, the current way the test is written, it would succeed even if the password were not generated, because Chrome changed since the introduction of the test in https://chromiumcodereview.appspot.com/23533069, and computes best matches separately by username.

On the other hand, there seems to be no special treating of generated passwords during the selection for best matches, so if there were 2 results for the same username, one of them scoring high but not generated, and the other scoring low but generated, it seems that Chrome would fill the high-scoring result, and not the generated one.

@dvadym -- assigning to you as the author of the current best-matching scheme, and the lead of generation. What is your opinion? Should we delete TestForceInclusionOfGeneratedPasswords_NoMatch? Is the current functionality covered by TestBestCredentialsByEachUsernameAreIncluded, or do we need to test more?

I am currently revising all PasswordFormManagerTests because of my work on FormFetcher, so it would be easiest for me to include any changes to these tests in my work. But I'd like to speak to you about this, so filing this bug so that we don't forget :).
 

Some PasswordFormManagerTest test cases seem out of date

=== TestForceInclusionOfGeneratedPasswords_NoMatch ===
The test verifies that saved passwords always are sent to be autofilled if they have the "generated" bit set. However, the current way the test is written, it would succeed even if the password were not generated, because Chrome changed since the introduction of the test in https://chromiumcodereview.appspot.com/23533069, and computes best matches separately by username.

On the other hand, there seems to be no special treating of generated passwords during the selection for best matches, so if there were 2 results for the same username, one of them scoring high but not generated, and the other scoring low but generated, it seems that Chrome would fill the high-scoring result, and not the generated one.

Should we delete TestForceInclusionOfGeneratedPasswords_NoMatch? Is the current functionality covered by TestBestCredentialsByEachUsernameAreIncluded, or do we need to test more?

=== CorrectlyUpdatePasswordsWithSameUsername ===
This test creates a situation where two logins with the same unique key are saved. That's not realistic. The introducing CL https://codereview.chromium.org/431893003 and the associated bug do not seem very clear about what was the original failure seen in reality. It seems that the test needs adjusting to work with logins with the same username but different unique keys.