ILSpy incorrectly disassembles System.Buffers.Binary.BinaryPrimitives to empty branches about ilspy

[siegfriedpammer]

If you add the instruction bytes to the IL output, (The setting can be found in Settings > Display > Decompilation view options > last checkbox).

.method /* 06003173 */ public hidebysig static 
	int32 ReadInt32BigEndian (
		valuetype System.ReadOnlySpan`1<uint8> source
	) cil managed aggressiveinlining 
{
	// Method begins at RVA 0x64c5d4
	// Header size: 1
	// Code size: 19 (0x13)
	.maxstack 8

	/* 0x0064C5D5 7E78030004         */ IL_0000: ldsfld bool System.BitConverter::IsLittleEndian /* 04000378 */
	/* 0x0064C5DA 2D00               */ IL_0005: brtrue.s IL_0007

	/* 0x0064C5DC 02                 */ IL_0007: ldarg.0
	/* 0x0064C5DD 284704002B         */ IL_0008: call !!0 System.Runtime.InteropServices.MemoryMarshal::Read<int32>(valuetype System.ReadOnlySpan`1<uint8>) /* 2B000447 */
	/* 0x0064C5E2 2865310006         */ IL_000d: call int32 System.Buffers.Binary.BinaryPrimitives::ReverseEndianness(int32) /* 06003165 */
	/* 0x0064C5E7 2A                 */ IL_0012: ret
} // end of method BinaryPrimitives::ReadInt32BigEndian

You will see that the bytes of the instruction are 0x2D00, where 0x2D is the opcode for brtrue.s:

Taken from ECMA-335, 6th edition, June 2012, page 297

and 0x00 is the relative offset to jump to, which basically is "don't jump at all".

Not sure what the C# compiler is trying to do, but it seems pointless. I don't think ILSpy is at fault here. Probably some branch optimization gone wrong in the compiler.

[siegfriedpammer]

Converter.IsLittleEndian is marked with an IntrinsicAttribute, so it seems there is something special going on. May I ask you to open a discussion on the dotnet/runtime repo?