I am currently testing the webtransport client impl against quiche based server (in fact it is also a node.js UDP on the other side).
The server is started, and on the client side, multiple clients are created in a row and also closed. (The different tests).
After 5 successful connects and sending data back and forth,
the 6 connect fails (after a few shuffles, it does not seem to matter what the actual test is doing).
(It is on Windows 11, it may matter).
I have traced it and the UDP packet of the 6th attempt is never received on the server side.
Though I can see with printf's that the WSASendTo is called and succeeds.
(Btw. it is not the first packet of a client connect, it is the one initiated by the sendHeadersafter creating a bidirectional stream).
It is not uncommon, that WSA causes troubles reusing ports. (Though the client endpoint is recycled, but it also appears, if I turn off the recycling).
The failing packet sending is a batched send. But also the single one after this are not captured on the receiving side.
Though I was not able to replicate it in a generic test.
@jasnell @pimterry Any ideas where to look? (For debug code...).
I am currently testing the webtransport client impl against quiche based server (in fact it is also a node.js UDP on the other side).
The server is started, and on the client side, multiple clients are created in a row and also closed. (The different tests).
After 5 successful connects and sending data back and forth,
the 6 connect fails (after a few shuffles, it does not seem to matter what the actual test is doing).
(It is on Windows 11, it may matter).
I have traced it and the UDP packet of the 6th attempt is never received on the server side.
Though I can see with printf's that the
WSASendTois called and succeeds.(Btw. it is not the first packet of a client connect, it is the one initiated by the
sendHeadersafter creating a bidirectional stream).It is not uncommon, that WSA causes troubles reusing ports. (Though the client endpoint is recycled, but it also appears, if I turn off the recycling).
The failing packet sending is a batched send. But also the single one after this are not captured on the receiving side.
Though I was not able to replicate it in a generic test.
@jasnell @pimterry Any ideas where to look? (For debug code...).