rgw: make sending Content-Length in 204 and 304 responses controllable

[rzarzynski]

No description provided.

[wido]

Always nasty to violate RFCs, but understandable. In which situation is this required?

Code looks good to me.

[rzarzynski]

@wido: thanks for the review. We need the configurable to pass RefStack due to the following test case:

tempest.api.object_storage.test_object_services.ObjectTest.test_delete_object

You might also want to take a look on the PR #9188. Anyway, I'm putting the DNM tag. There is a new hope for fixing the issue in Tempest.

[wido]

Ok, great! I wouldn't want to encourage people to violate RFCs, they are there for a reason.

Don't mind if this gets merged though.

[yehudasa]

@rzarzynski do we still need this?

[rzarzynski]

@yehudasa: fortunately no :-) as our fix in Tempest has been merged. I will push 93629c0 in a separated pull request. Closing this one.

[rzarzynski]

@mattbenjamin: I've just pushed a version ported to current master. DNM till testing on the morning.

[rzarzynski]

@mattbenjamin:

Manual testing:

• For rgw_print_prohibited_content_length = false (default):

$ curl -i "${publicURL}/cont/obj10" -X DELETE -H "X-Auth-Token: ${token}" 
HTTP/1.1 204 No Content
X-Trans-Id: tx00000000000000000000b-0058a6eb39-1018-default
Content-Type: text/plain; charset=utf-8
Date: Fri, 17 Feb 2017 12:23:21 GMT

• For rgw_print_prohibited_content_length = true:

$ curl -i "${publicURL}/cont/obj10" -X DELETE -H "X-Auth-Token: ${token}" 
HTTP/1.1 204 No Content
X-Trans-Id: tx000000000000000000010-0058a6ea8c-1015-default
Content-Type: text/plain; charset=utf-8
Content-Length: 0
Date: Fri, 17 Feb 2017 12:20:28 GMT

However, Tempest has found a regression in tempest.api.object_storage.test_container_services.ContainerTest.test_list_container_contents_with_end_marker. However, most likely it is unrelated to the change. I'm working on this.

[rzarzynski]

@mattbenjamin: definitely the regression in Tempest is unrelated to the change.

[rzarzynski]

Created a ticket to track the issue: http://tracker.ceph.com/issues/18977.

[mattbenjamin]

the change appears logically and trivially correct; it's 100/200 subset rgw suite run has failures which are clearly unrelated, merging to avoid schedule impact