I took at a look at the pcap file that is failing and I've tracked down the first two differences in the file. I'm not sure where the problem is in Daffodil, but this should help to narrow down the search.
The first packet is unparsed correctly. The second and third packets are not. The Options field in the TCPHeader of the second packet is: <Options>0204059804020000</Options> This is exactly what the data looks like in the original file starting at byte position 224 (0xEO in hex). When this blob is unparsed, it results in the following hex: 02 04 05 F8 04 02 00 00 Note the only difference here is the 4th byte which is 0xF8 instead of 0x98. So the upper nibble of that byte is 0b1111 instead of 0b1001. The type of Options is xs:hexBinary. The second incorrect byte is in the Checksum field of the third packet (byte position 318, 0x13E in hex). The correct value is (16-bit BE xs:unsignedInt) 22312 but is unparsed as 22520. Converted to hex these values are: 22312: 0x5728 22520: 0x57F8 So again, we see some upper nibble converted to 0b1111. It's not clear why. Running a binary diff on the correct output and the unparsed output shows the the differences are always some upper nibble (the value varies) being converted to 0b1111. So something in the unparse logic isn't handling this upper nibble correctly. In one case it was an xs:hexBinary blob, in the other case it was a 2-byte integer. So maybe the bug is in is some shared logic in unparsing hexBinary and integers, or perhaps maybe related to collapsing buffered DOS's to direct? It's not exactly clear, and it doesn't happen all that often (I think there's only 8 occurrences in this file). But when it does happen, it's always some upper nibble converted to 0b1111. I'll continue investigating, but hopefully this should narrow down where we need to be looking. - Steve This is just a hex binary blob of data. On 12/20/2017 07:21 PM, Mike Beckerle wrote: > So I have fixed all the issues with PR 12 save 1. > > > I added 4 commits. Granular is better with this review system I've determined. > > > Any insight into what changed that caused the need for the changes to > InteractiveDebugger.scala in the last (6th) commit... are welcome. I didn't > like putting this hack in. > > > Note: this branch is yet to be rebased onto master. I still have that todo. > > > Status is all tests pass, sbt cli all tests pass, all DFDLSchemas and fouo > schemas pass, except 1: > > > PCAP test_pcap_test_http_ipv6 still fails. > > > Steve L - can we use wireshark or some other tool to verify that the > checksums in the expected infoset XML for this test are in fact correct? > > I don't want to debug in detail if these were never verified before. > > > I still have these todos: > > * changes to make per review comments on the 2nd commit > * rebase onto master. > * PCAP test failure mystery > >
