Bug #3933
closed
Nack test incorrect counting of Nacks
100%
Description
In IntegrationTests Nack scenario test 3, two Nacks were returned from two upstream faces, and NFD is expected to return only one Nack to the downstream face.
This was implemented as parsing the log of the consumer application and expecting the consumer application to receive only one Nack.
Since #3908, after a consumer expresses multiple Interests with same Name+Selectors+Link, when a Nack with specified Name+Selectors+Link arrives, all pending Interests would be Nacked, breaking the test assumption. Furthermore, in case NFD incorrectly returns more than one Nack, excessive Nacks received by ndn::Face would be dropped.
Thus, the test should be re-implemented by counting the number of Nacks between router NFD and consumer NFD (A-D), through parsing of ndndump or tcpdump logs.
Files
Updated by Junxiao Shi about 7 years ago
- File 20170125191715.tgz 20170125191715.tgz added
- Assignee set to Hila Ben Abraham
Test log is attached.
This is assigned to Hila who implemented this scenario.
Updated by Hila Ben Abraham about 7 years ago
Thus, the test should be re-implemented by counting the number of Nacks between router NFD and consumer NFD (A-D), through parsing of ndndump or tcpdump logs.
If I understand it correctly, steps #10 and #12 in #3157 should be changed to expect D to receive exactly two Nacks, one for each nonce sent. In addition, we should check that no additional nacks were incorrectly sent, and dropped by ndn::Face, by checking the actual packets sent using ndndump.
Am I right? Is it possible to reopen #3157 with an up-to-date version of the test's steps?
Updated by Junxiao Shi about 7 years ago
Reply to note-2:
No, the test assertions (expected number of Nacks) are correct. However, the actual number of Nacks cannot be counted from the consumer application, but should be counted on the network link by parsing ndndump logs.
Updated by Junxiao Shi almost 7 years ago
- Assignee changed from Hila Ben Abraham to Eric Newberry
Eric volunteered to fix all broken integ on 20170419 call.
Updated by Eric Newberry almost 7 years ago
- Related to Bug #4058: nfdc input and output format changes cause integration tests to fail added
Updated by Eric Newberry almost 7 years ago
- Status changed from New to In Progress
Updated by Eric Newberry almost 7 years ago
Junxiao Shi wrote:
Reply to note-2:
No, the test assertions (expected number of Nacks) are correct. However, the actual number of Nacks cannot be counted from the consumer application, but should be counted on the network link by parsing
ndndumplogs.
So therefore changes should be made to both test 3 and test 4?
Updated by Eric Newberry almost 7 years ago
- Status changed from In Progress to Code review
- % Done changed from 0 to 100
Updated by Eric Newberry almost 7 years ago
- Status changed from Code review to Closed