-
Notifications
You must be signed in to change notification settings - Fork 2k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Sapier's fix for the RESEND RELIABLE problem #4170
Conversation
👎 |
That's precisely what's happening here, read the code. |
And, unless you do exactly that, the log fills up with gigabytes, literally, of RESEND RELIABLE messages and worlds go down. Without this patch, some server owners are locked out of Minetest entirely. One of them, in late 2014, was me. The project would have lost a dozen worlds and other contributions without this patch. |
Putting the word "fix" into quotes, @kwolekr, so as to deprecate its status as such, does not change any of what I've said. |
D'oh, you're right. Okay then. |
Generally I like the change as it makes the server more robust to rogue client behaviour. But I do think that the underlying issue deserves more attention. It is not just simple network weirdness which would cause this, it can be either rogue or buggy clients, or a bug in the server somewhere. Look at #4138 for example, it has caused this very RESEND_RELIABLE itself as well (see commit msg of 423d8c1 for more info). |
I'm okay with this solution as a hacky patchy workaroud because let's face it, connection.cpp is a mess. I think our manhours are best spent making something that's unit-tested for edge cases, better documented, and easier to maintain. If anybody wants to try tracking down the root cause, that's cool too, so we've decided to wait 2 weeks before merging this. But I personally 👍 this, provided the obnoxious style errors are fixed before merging. |
It would be cool for some contributor to track it down. |
We NEED this 'fix' as many server owners are frustrated with their worlds getting stuck and no way to reset server unless they are physically there to do so. |
You can't believe how this issue is annoying to admins and players, it destroys gameplay, totally.
I'm not sure if this is related to this patch, but this can give some clues about what is broken in network code. Maybe it was related to ESM only, that has lots of mods. |
|
This is an important first step. However, tested 👍 |
Well as @tenplus1 is in desparate need for the fix and the chance that the original cause for this bug will be found by an external contributor is near zero, I now think that there should be no 2 week stall anymore. Its a bit sad for the engine, as the original cause won't be found, but xanadu is a great server so why not support it. |
@est31 - I'm sure that through time the cause will be found, but for now stopping it from breaking servers is more important. |
👍 |
We lose players because of this problem, two weeks delay is not good. |
👍 as workaround. |
This is Sapier's fix for the RESEND RELIABLE problem, a network issue that has affected MT worlds for at least 1.5 years. The fix is relatively simple and is believed to be correct.
I've used the fix myself for over a year to keep a dozen worlds up and running; for me, MT wouldn't be possible without it. The fix has been tested, recently, in other worlds and seems to improve performance even in problematic cases such as residential hosting.