Memory leak in readable-stream when sending packets

[IceTank]

[x] The FAQ doesn't contain a resolution to my issue

Versions

• minecraft-protocol: latest
• server: vanilla/spigot/paper 1.12.2
• node: 14.x

Detailed description of a problem

Writing packets on a nmp server can cause a memory leak the serializer. This is what my heap looks like after running out off memory

Object×979172	2	54 545 95253 %	85 308 40082 %
Client×3	9	6720 %	72 301 20870 %
Client@1661959	15	2240 %	52 042 94850 %
serializer::Serializer@1666511	16	1280 %	49 284 12048 %
_writableState::WritableState@1949265	17	2400 %	49 282 96048 %
bufferedRequest::Object@1765163	18	640 %	49 282 34448 %
next::Object@1764833	19	640 %	49 282 12848 %
next::Object@1764837	20	640 %	49 281 91248 %
next::Object@1764767	21	640 %	49 281 69648 %
next::Object@1763723	22	640 %	49 281 48048 %
next::Object@1763727	23	640 %	49 281 26448 %

The next go on for some time as they represent the queued up data in the Transform that is causing the memory leak.

I am not sure what might be causing this. But I think this is caused by protodef trying to parse invalid packets and failing. I had a similar issue in mineflayer where nmp would completely brick itself when it was trying to send invalid sign update packets. My guess is that something messes up causing the internal logic to break so packets are not parsed correctly and end up clogging up the buffer with garbage data.

Expected behavior

No memory leak

[extremeheat]

How long does it take to get this error & what server is it on?

Did you dump the raw inbound packets also? This will make it easier to debug

[IceTank]

How long does it take to get this error & what server is it on?

Did you dump the raw inbound packets also? This will make it easier to debug

This happened after about 8-12 hours. The server is 2b2t. I think this might have something to do with the fact that that server sends a login packet after you left the queue to log you in to the main server. My guess is that there is something somewhere not realizing that some destination is unavailable anymore. I was not able to reproduce this issue with local test yet. So it might be my code that is wrong?

[extremeheat]

So it might be my code that is wrong?

Yes probably if you cannot reproduce with a basic code sample

[IceTank]

I never really found what was causing this but I am 80% sure that it was some weird interaction with an infinity loop I had in my code that was blocking the event loop