Conversation
| _verify_atomic_key_data(bucket_name, objname, file_size, "B") | ||
|
|
||
|
|
||
| @pytest.mark.fails_on_dbstore |
There was a problem hiding this comment.
What are the test expectations here? Why is it expecting any particular order?
There was a problem hiding this comment.
I'm not sure about test expectations, but I see the tests write in parallel the object with the same name
There was a problem hiding this comment.
Looks like the test is a flawed, but can be fixed. It says
confirm that it is all one or the other
But this means that any of the results (A or B) is correct. And the way it works is uploading objects ~concurrently, so we can't rely on B always being the last one. So to me that's the way to fix this test, not by disabling it (I admit it's a bit strange one, but the idea is to never get an AB mix which can theoretically happen).
There was a problem hiding this comment.
Test initiates one object upload, during this upload it starts and finishes another upload of the same object and after that the test finishes the first upload. Test expectation is that the object will have the contents of the first upload, because it finishes later.
But the test relies on the implementation of the boto3 put_object and I'm not really sure if the first upload really finishes later. But, if the test worked previously, maybe it was a correct assumption. Otherwise it would be a big flipper.
There was a problem hiding this comment.
If we expect rewind_put_fp_a() to finish first we can also add a delay into it.
smallhive
force-pushed
the
adjust-tests
branch
2 times, most recently
from
673df17 to
856cc82
Compare
smallhive
force-pushed
the
adjust-tests
branch
2 times, most recently
from
42fee77 to
ac6b811
Compare
This delay is a temporary solution. It required to be able to get the right order of objects. There is some limitation to understand what object was first in the same second. See nspcc-dev/neofs-s3-gw#1066. Signed-off-by: Evgenii Baidakov <evgenii@nspcc.io>
These tests write twice object with the same name. There is some limitation to understand what object was first in the same second. See nspcc-dev/neofs-s3-gw#1066. Signed-off-by: Evgenii Baidakov <evgenii@nspcc.io>
3 participants
No description provided.