panic: protobuf tag not enough fields in Message.state

[dennis-tra]

I updated Nebula dependencies last week and decoding protobuf messages from other peers panic:

caught panic: protobuf tag not enough fields in Message.state: 
goroutine 170 [running]:
runtime/debug.Stack()
        /Users/.../go/src/runtime/debug/stack.go:26 +0x64
github.com/libp2p/go-msgio/protoio.(*uvarintReader).ReadMsg.func1()
        /Users/.../.cache/go/pkg/mod/github.com/libp2p/go-msgio@v0.3.0/protoio/uvarint_reader.go:65 +0x40
panic({0x10673b400?, 0x14001364dd0?})
        /Users/.../go/src/runtime/panic.go:785 +0x124
github.com/gogo/protobuf/proto.(*unmarshalInfo).computeUnmarshalInfo(0x140004c1220)
        /Users/.../.cache/go/pkg/mod/github.com/gogo/protobuf@v1.3.2/proto/table_unmarshal.go:341 +0xfa0
github.com/gogo/protobuf/proto.(*unmarshalInfo).unmarshal(0x140004c1220, {0x106a7c160?}, {0x140012c4000, 0xc49, 0xc49})
        /Users/.../.cache/go/pkg/mod/github.com/gogo/protobuf@v1.3.2/proto/table_unmarshal.go:138 +0x4c
github.com/gogo/protobuf/proto.(*InternalMessageInfo).Unmarshal(0x107b105f0?, {0x106b63ca0, 0x1400071e600}, {0x140012c4000?, 0xc49?, 0xc49?})
        /Users/.../.cache/go/pkg/mod/github.com/gogo/protobuf@v1.3.2/proto/table_unmarshal.go:63 +0xc4
github.com/gogo/protobuf/proto.(*Buffer).Unmarshal(0x14001133e90, {0x106b63ca0, 0x1400071e600})
        /Users/.../.cache/go/pkg/mod/github.com/gogo/protobuf@v1.3.2/proto/decode.go:424 +0x174
github.com/gogo/protobuf/proto.Unmarshal({0x140012c4000, 0xc49, 0xc49}, {0x106b63ca0, 0x1400071e600})
        /Users/.../.cache/go/pkg/mod/github.com/gogo/protobuf@v1.3.2/proto/decode.go:342 +0x10c
github.com/libp2p/go-msgio/protoio.(*uvarintReader).ReadMsg(0x14001abe240, {0x106b63ca0, 0x1400071e600})
        /Users/.../.cache/go/pkg/mod/github.com/libp2p/go-msgio@v0.3.0/protoio/uvarint_reader.go:85 +0x1a4
github.com/dennis-tra/nebula-crawler/libp2p.ctxReadMsg.func1({0x106b60208?, 0x14001abe240?})
        /Users/.../nebula/libp2p/rpc.go:54 +0x68
created by github.com/dennis-tra/nebula-crawler/libp2p.ctxReadMsg in goroutine 1817
        /Users/.../nebula/libp2p/rpc.go:52 +0xec

Downgrading to github.com/libp2p/go-libp2p-kad-dht to v0.28.2 and as a consequence also github.com/libp2p/go-libp2p-record to v0.2.0 (v0.3.0 didn't work) solved the problem. Here's the roll-back commit.

I presume it's related to the recent protobuf update: #975

It panics when I do the following:

import (
	"github.com/libp2p/go-msgio/protoio"
	pb "github.com/libp2p/go-libp2p-kad-dht/pb"
)

var r protoio.ReadCloser
var mes *pb.Message

r.ReadMsg(mes) // <--- panics here

Does the change require me to decode messages differently?

[MarcoPolo]

I've not yet looked closely at this, but at a glance I believe it may be a local issue where Nebula is using the go-msgio/protoio package intended for use with gogo protobuf types for a type that is not gogo protobuf (instead it is now the Google protobuf type).

[guillaumemichel]

The fix on Nebula should be dennis-tra/nebula#88.

@dennis-tra you can provide an update here after trying the fix.