Closed fercasgal closed 9 months ago
@fercasgal I pushed out #416, which should fix this. If we get into handle_error/2
, it means that there was a transport error, and I think we assumed that that meant the socket was closed.
I tried your branch in our setup and it fixed the memory leak we were seeing.
Fantastic! I've released v1.5.2.
When there's a receive timeout, the
Mint.HTTP1
simply does:But the socket is still open and the calling code has no chance of closing it, since:
So the calling code believes that it has to open a new connection, and the previous one leaks.