Broken test/test_record_padding: I/O issues cause non zero exit sometimes

[johubertj]

Problem:

Reference : #5099

test_record_padding has been changed to expect a stderr and a non zero exit code (above PR). This needs more investigation if this is the expected behavior. Without the change the test result is flaky. There is a different ratio of success/fail/skip every time it's ran. This is most likely due to I/O loop.

This was also tested with openSSL 3. Other test cases that have a similar setup (s2n is a client and there is a close marker)

• test_key_update.py PASSES
• test_renegotiate.py PASSES
• test_serialization.py FAILS

Solution:

Solution is unknown and needs more investigation

[maddeleine]

Can you explain the significance of the 3 integ tests you mentioned passing/failing here? They pass in CI, so I'm not sure why that note is relevant.

[johubertj]

Can you explain the significance of the 3 integ tests you mentioned passing/failing here? They pass in CI, so I'm not sure why that note is relevant.

We fixed test_record_padding by assuming we should receive a stderr and a non zero output. The 3 above integ tests have a similar set up to test_record_padding but do not assume that they should receive a stderr and a non zero output. This seems odd! If multiple test cases have the same setup, they should probably also have the same output.