run-v1model.analysis.md

A. Fix Logic

1. Typecheck error

• table-entries-ser-enum-bmv2.p4

Refer to A.1.(3) Implicit cast involving serializable enums in typechecker error analysis.

2. Stf parser error

• ipv6-switch-ml-bmv2.stf

B. Need Test Clarification

1. Payload is not extracted

[FAIL] Expected: 0 0000000000000000000000000000
       Got:      0 0000000000000000000000000000000000

The trailing 3 bytes should be the payload, i.e., should not be extracted by the P4 program. The program defines a header

struct Headers {
    ethernet_t eth_hdr;
    H h;
}

and only extracts the eth_hdr header:

pkt.extract(hdr.eth_hdr);

Adding pkt.extract(hdr.h); would extract the h header and the test would pass.

• gauntlet_invalid_hdr_short_circuit-bmv2.p4

2. Header is not emitted

expect 2 00000101 ******** **** 7f 66
// expect 2 00000101 ******** **** 7f 66 **** ** **
packet 0 00000101 00000202 0303 55 66 7777 88 00

The test does not expect the header stack to be emitted. After patching the SFT test as in the comment, the test would pass.

• ternary2-bmv2.stf

C. Unsupported Features

1. Table priority annotation

const entries = {
    0x1111 &&& 0xF    : a_with_control_params(1) @priority(3);
    0x1181            : a_with_control_params(2);
    0x1181 &&& 0xF00F : a_with_control_params(3) @priority(1);
}

Specifying entry priority via @priority annotation is not supported in p4cherry. But patching the test as below would make it pass:

largest_priority_wins = false;
entries = {
    priority=3: 0x1111 &&& 0xF    : a_with_control_params(1);
                0x1181            : a_with_control_params(2);
    priority=1: 0x1181 &&& 0xF00F : a_with_control_params(3);
}

• table-entries-priority-bmv2