[test cases] igmp generation

[zeeshanlakhani]

Test cases for IGMP v2/v3 generation.

[zeeshanlakhani]

@FelixMcFelix: This is currently in draft status, as I wanted to chat about some validation pieces around the multicast address range and a few other fields. We'll take on validation in #2, separately.

[FelixMcFelix]

Thanks Zeeshan -- the IGMP tests look good. There are a few nits around the added methods for IP addresses. I think in general we should be trying to match constness, naming, and matched IP ranges to what is done in core::net?

I was thinking about this since our From/Into are zero-cost there, and realised that in core every check is done on self.octets() rather than reaching through the ref. If we do the same we generate identical code to just inlining the equivalent core method:

/// AA
#[no_mangle]
pub const fn is_documentation(ip: &Ipv4Addr) -> bool {
    matches!(ip.inner, [192, 0, 2, _] | [198, 51, 100, _] | [203, 0, 113, _])
}

/// AA
#[no_mangle]
pub const fn is_documentation2(ip: &Ipv4Addr) -> bool {
    let ip = *ip;
    matches!(ip.inner, [192, 0, 2, _] | [198, 51, 100, _] | [203, 0, 113, _])
}

/// B
#[no_mangle]
pub const fn is_documentation3(ip: &Ipv4Addr) -> bool {
    let ip = core::net::Ipv4Addr::new(ip.inner[0], ip.inner[1], ip.inner[2], ip.inner[3]);
    ip.is_documentation()
}

`cargo asm` output. `is_documentation2` and `is_documentation3` are identical.

kyle@Mac oxpacket % cargo asm -p ingot-types
    Finished `release` profile [optimized + debuginfo] target(s) in 0.01s

Try one of those by name or a sequence number
 0 "<&T as core::fmt::Display>::fmt" [14, 57]
 2 "<ingot_types::error::CRStr as core::fmt::Debug>::fmt" [52]
 3 "<ingot_types::error::CRStr as core::fmt::Display>::fmt" [52]
 4 "<ingot_types::error::CRStrError as core::fmt::Display>::fmt" [18]
 5 "<ingot_types::error::PacketParseError as core::fmt::Display>::fmt" [64]
 6 "<ingot_types::error::ParseError as core::fmt::Display>::fmt" [130]
 7 "ingot_types::error::CRStr::new_unchecked::panic_cold_explicit" [15]
 8 "is_documentation" [67]
 9 "is_documentation2" [32]
10 "is_documentation3" [32]
kyle@Mac oxpacket % cargo asm -p ingot-types 8
    Finished `release` profile [optimized + debuginfo] target(s) in 0.01s

	.globl	_is_documentation
	.p2align	2
_is_documentation:
Lfunc_begin7:
	.cfi_startproc
	ldrb w8, [x0]
	cmp w8, #203
	b.eq LBB7_7
	cmp w8, #198
	b.eq LBB7_5
	cmp w8, #192
	b.ne LBB7_8
	ldrb w8, [x0, #1]
	cbz w8, LBB7_10
	mov w0, #0
	ret
LBB7_5:
	ldrb w8, [x0, #1]
	cmp w8, #51
	b.ne LBB7_8
	ldrb w8, [x0, #2]
	cmp w8, #100
	cset w0, eq
	ret
LBB7_7:
	ldrb w8, [x0, #1]
	cbz w8, LBB7_9
LBB7_8:
	mov w0, #0
	ret
	ldrb w8, [x0, #2]
	cmp w8, #113
	cset w0, eq
	ret
	ldrb w8, [x0, #2]
	cmp w8, #2
	cset w0, eq
	ret
kyle@Mac oxpacket % cargo asm -p ingot-types 9
    Finished `release` profile [optimized + debuginfo] target(s) in 0.01s

	.globl	_is_documentation2
	.p2align	2
_is_documentation2:
Lfunc_begin8:
	.cfi_startproc
	ldrb w8, [x0]
	ldrb w9, [x0, #1]
	ldrb w10, [x0, #2]
	cmp w9, #0
	mov w11, #113
	ccmp w10, w11, #0, eq
	cset w11, eq
	cmp w9, #51
	mov w12, #100
	ccmp w10, w12, #0, eq
	cset w12, eq
	cmp w9, #0
	ccmp w10, #2, #0, eq
	cset w9, eq
	cmp w8, #192
	csel w9, wzr, w9, ne
	cmp w8, #198
	csel w9, w12, w9, eq
	cmp w8, #203
	csel w0, w11, w9, eq
	ret
kyle@Mac oxpacket % cargo asm -p ingot-types 10
    Finished `release` profile [optimized + debuginfo] target(s) in 0.01s

	.globl	_is_documentation3
	.p2align	2
_is_documentation3:
Lfunc_begin9:
	.cfi_startproc
	ldrb w8, [x0]
	ldrb w9, [x0, #1]
	ldrb w10, [x0, #2]
	cmp w9, #0
	mov w11, #113
	ccmp w10, w11, #0, eq
	cset w11, eq
	cmp w9, #51
	mov w12, #100
	ccmp w10, w12, #0, eq
	cset w12, eq
	cmp w9, #0
	ccmp w10, #2, #0, eq
	cset w9, eq
	cmp w8, #192
	csel w9, wzr, w9, ne
	cmp w8, #198
	csel w9, w12, w9, eq
	cmp w8, #203
	csel w0, w11, w9, eq
	ret

Long way of saying I'm not averse to us doing a const conversion and calling core's methods.

Some of these methods are also kinda hard to define without a prefix length, which feels like we're retreading a lot of the work that went into oxnet. It's a bit sad that the zerocopy traits aren't on the core/std IpAddr types for it all to just work together, but I guess if we want to go between it should be free.

[FelixMcFelix]

Two nits, but otherwise looks good. It's kind of a shame that from_octets isn't yet stable in core, or this would be neater still. Thanks!