Container arguments with type infint are treated as type variables

[mihaibudiu]

Fixes #3886

The problem is the following: given a constructor like control C()(int x), the type of x is unified with the argument in every invocation of the constructor. But not all invocations must have the exact same argument type.

In other parts of the compiler type 'int' (Type_InfInt) is treated like a type variable - and unification is used to decide the actual concrete type of an int expression. Here we do the same: when the control C is instantiated, we treat the type of x as a fresh type variable; in this way, the arguments of two separate instantiations of C do not need to unify with each other (via the common constructor type).

C(5) c1;
C(4) c2;

We do not require 4 and 5 to have the exact same type, they are different int types.

[mihaibudiu]

@fruffy this also fixes some problems with the README, please review.

[mihaibudiu]

This should supersede #3884

[fruffy]

Updating the libbpf dependency might cause issues with the psa-ebpf or ebpf-kernel repository? Does that happen? I guess the XDP one is deprecated.

[fruffy]

We should probably reduce our use of boost now that we have C++17 support.

[ChrisDodd]

👍

[ChrisDodd]

Looks good, and fixes my testcases. If you check this in, I'll prep a change adding just the test.

[ChrisDodd]

👍

[ChrisDodd]

Updating the libbpf dependency might cause issues with the psa-ebpf or ebpf-kernel repository? Does that happen? I guess the XDP one is deprecated.

Looks like this is actually an unrelated change that crept into this PR?

[mihaibudiu]

I have squashed the commits and added tests.

[mihaibudiu]

Rebased on main to pickup the macos fixes so we can merge.

[mihaibudiu]

@fruffy this fails with an assertion failure in the testing module.

What can we do to merge this PR?

 734/1528 Test #3373: testgen-p4c-bmv2-ptf/issue2345.p4 .......................................***Failed    2.79 sec
============ Program seed 1000 =============

warning: Ingress parser exception handler not fully implemented
============ Program trace for Test 0 ============

p: [Parser] p
[State] start
[Event]: Extract: Succeeded
[Extract] hdr.eth_hdr.dst_addr; = 0x0000_0000_0000
[Extract] hdr.eth_hdr.src_addr; = 0x0000_0000_0000
[Extract] hdr.eth_hdr.eth_type; = 0x0000
[Event]: Extract Condition: p4t*zombie.const.0.*packetLen_bits >= 112; | Extract Size: 112
[State] accept
[Event]: Control vrfy start
[Event]: Control ingress start
[Expression]: If Statement false: 2 == 1; = 0;
[Expression]: If Statement false: 0; = 0;
[Expression]: If Statement true: 1; = 1;
[Expression]: If Statement true: 1; = 1;
[Expression]: If Statement false: 0 != 0; = 0;
[Event]: Control egress start
[Event]: Control update start
[Event]: Control deparser start
[Emit] h.eth_hdr.dst_addr; = 0x0000_0000_0002
[Emit] h.eth_hdr.src_addr; = 0x0000_0000_0001
[Emit] h.eth_hdr.eth_type; = 0x0002
[Emit] h.eth_hdr.*valid; = 1;
[Event]: Prepending the emit buffer to the program packet.
[Expression]: If Statement false: 511 == 0; = 0;
=======================================
============ Input packet for Test 0 ============
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
=======================================
Input packet size: 2288
============ Ports for Test 0 ============
Input port: 1
=======================================
============ Output packet for Test 0 ============
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
=======================================
Output packet size: 2288 
=======================================
============ Output mask Test 0 ============
FFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF
=======================================
Output port: 0

=======================================
============ Test 0: Statements covered: 0.888889 (8/9) ============
============ End Test 0 ============

[fruffy]

This could be non-determinism, let me rerun first.

[fruffy]

The test is passing now. Looks like a fluke because I do not understand how this line can fail.