looking for IfMatch/IfNoneMatch in s3-tests, i only see test cases for get_object() which wouldn't call through prepare_atomic_modification()

could you please add test cases for put_object() and complete_multipart_upload() that covers both etag values and the * wildcard?

looking for IfMatch/IfNoneMatch in s3-tests, i only see test cases for get_object() which wouldn't call through prepare_atomic_modification()

could you please add test cases for put_object() and complete_multipart_upload() that covers both etag values and the * wildcard?

sure, @mattbenjamin said that there are more cases that need fix such as:
multidelete and multiput looking for a LastModified
I will do that then write the tests

I think the issue in conditional put with versioned bucket, is this block of code in RGWRados::get_obj_state_impl:

  if (s->has_attrs) {
    if (s->is_olh && need_follow_olh) {
      return get_olh_target_state(dpp, *octx, bucket_info, obj, s, psm, y);
    }
    return 0;
  }

we return from this function and the object doesn't have an etag in attrs

why are we expecting an attrs here?

I think the issue is that there are two objects in question here. The unversioned object, and the versioned object. That sequence you posted does load the state (and so the attributes) of the versioned object, but it doesn't do anything to the unversioned object. Where is this being called from?

the subject says "fix conditional write" but is this PR going to address DeleteObject only?

I think the issue in conditional put with versioned bucket, is this block of code in RGWRados::get_obj_state_impl:

#60957 reported a similar issue here, but the fix wasn't passing tests :(

edit: i think that was specific to multipart uploads though

I think the issue in conditional put with versioned bucket, is this block of code in RGWRados::get_obj_state_impl:

@AliMasarweh what are you expecting conditional put to do in a versioned bucket? with versioning enabled, writes create a new object version instead of overwriting an existing one. so If-Match would not find an existing etag for comparison

edit: maybe i'm mistaken, https://docs.aws.amazon.com/AmazonS3/latest/userguide/conditional-writes.html#conditional-error-response talks about interactions with versioning enabled/suspended

https://pulpito.ceph.com/alimasa-2025-07-17_02:35:01-rgw-wip-alimasa-68183-distro-default-smithi/
is this run good enough?

https://pulpito.ceph.com/alimasa-2025-07-17_02:35:01-rgw-wip-alimasa-68183-distro-default-smithi/ is this run good enough?

yikes, lots of red there. looks mostly unrelated except for one rgw/verify job: https://qa-proxy.ceph.com/teuthology/alimasa-2025-07-17_02:35:01-rgw-wip-alimasa-68183-distro-default-smithi/8392360/teuthology.log

test_s3.py::test_delete_objects_version_if_match_last_modified_time FAILED [ 94%]

___________ test_delete_objects_version_if_match_last_modified_time ____________

    @pytest.mark.fails_on_aws # only supported for directory buckets
    @pytest.mark.conditional_write
    def test_delete_objects_version_if_match_last_modified_time():
        client = get_client()
        bucket = get_new_bucket(client)
        check_configure_versioning_retry(bucket, "Enabled", "Enabled")
        key = 'obj'

        badmtime = datetime.datetime(2015, 1, 1)

        version = client.put_object(Bucket=bucket, Key=key)['VersionId']
        client.put_object(Bucket=bucket, Key=key)
        mtime = client.head_object(Bucket=bucket, Key=key)['LastModified']

        response = client.delete_objects(Bucket=bucket, Delete={'Objects': [{'Key': key, 'VersionId': version, 'LastModifiedTime': badmtime}]})
        assert 'PreconditionFailed' == response['Errors'][0]['Code']

        response = client.delete_objects(Bucket=bucket, Delete={'Objects': [{'Key': key, 'VersionId': version, 'LastModifiedTime': mtime}]})
>       assert key == response['Deleted'][0]['Key'] # success
E       KeyError: 'Deleted'

s3tests_boto3/functional/test_s3.py:18385: KeyError

do we have a stable commit on main that I should rebase on my code, teuthology has too many valgrind error:
https://pulpito.ceph.com/alimasa-2025-07-21_14:44:30-rgw-wip-alimasa-68183-distro-default-smithi/

qa/rgw/s3tests-branch.yaml modification should be cleaned up as well
#63348 (comment)

jenkins test make check arm64

This PR appears to have caused some significant breakage:

• https://tracker.ceph.com/issues/72398
• https://tracker.ceph.com/issues/72517

A new error regarding orphan list showed up during the teuthology run that indicated a regression:

https://pulpito.ceph.com/alimasa-2025-07-28_14:08:21-rgw-wip-alimasa-68183-distro-default-smithi/8412051

I bisected the orphan list failure to commit 55f5b76 , which is one of the two commits in this PR.

Do you have any plans to backport this to squid?

Do you have any plans to backport this to squid?

The tracker does not list any backports, and we try to record backports there. But @AliMasarweh may know differently....

Do you have any plans to backport this to squid?

thanks @cupnes, i updated https://tracker.ceph.com/issues/68183 to request squid and tentacle backports

@ivancich @cbodley
I understood. Thank you!