The t's discuss from zllovesuki

t's Issues

Improve peer graph implementation

panic: runtime error: index out of range [3] with length 3

goroutine 513 [running]:
github.com/zllovesuki/t/state.(*PeerGraph).removePeer(0xc000376f30, 0x2a0df27b1d5ea54a)
        /home/runner/work/t/t/state/peer_graph.go:80 +0x378
github.com/zllovesuki/t/state.(*PeerGraph).RemovePeer(0xc000376f30, 0x2a0df27b1d5ea54a)
        /home/runner/work/t/t/state/peer_graph.go:70 +0x4e
github.com/zllovesuki/t/server.(*Server).removePeer(0xc000276a00, 0x2a0df27b1d5ea54a)
        /home/runner/work/t/t/server/peers.go:197 +0x185
github.com/zllovesuki/t/server.(*Server).handlePeerEvents.func4(0xc000276a00, {0xe66d28, 0xc0000f3ce0})
        /home/runner/work/t/t/server/peers.go:180 +0xfa
created by github.com/zllovesuki/t/server.(*Server).handlePeerEvents
        /home/runner/work/t/t/server/peers.go:176 +0x1be

Consider using a kv instead of spare matrix

server: reconnect to peer if gossip is still reporting as alive

currently if error occurred during stream accept routine, the connection with the peer is dropped silently. however, it could be due to network blip, and gossip is still reporting the peer as alive. this edge case should be handled

Recommend Projects