Page 1 of 1
Not desync safe!
Posted: Sun Mar 11, 2018 3:35 am
by Mylon
So I was trying to play some multiplayer with a modpack using RSO 3.5.7 when it started desyncing like crazy.
Using the trusty power of meta tables, I added this to control.lua and started a fresh map:
Code: Select all
--Time for the debug code. If any (not global.) globals are written to at this point, an error will be thrown.
--eg, x = 2 will throw an error because it's not global.x or local x
setmetatable(_G, {
__newindex = function(_, n, v)
log("Desync warning: attempt to write to undeclared var " .. n)
-- game.print("Attempt to write to undeclared var " .. n)
global[n] = v;
end,
__index = function(_, n)
return global[n];
end
})
And as a result, I got this in the log file:
There's a lot of vars not properly being declared as local!
Re: Not desync safe!
Posted: Sun Mar 11, 2018 5:22 pm
by orzelek
Do you have any idea why it would suddenly become a problem?
It's been like that from quite some time and there were no desync reports. So either people are not reporting the desyncs or there is something else happening.
I can fix those but I'm not sure if it will help with your desync problems.
Re: Not desync safe!
Posted: Sun Mar 11, 2018 5:44 pm
by eradicator
I don't see why using globals would be any more or less desync prone than using locals. Care to explain?
Re: Not desync safe!
Posted: Sun Mar 11, 2018 5:50 pm
by orzelek
eradicator wrote:I don't see why using globals would be any more or less desync prone than using locals. Care to explain?
I think it's because global context is not synchronized in multiplayer - so any connecting client will have it clean. And local guarantees that variable won't be reused in other location (even by accident) since it won't exist there.
I did found an actual bug thanks to that debug code.
Re: Not desync safe!
Posted: Sun Mar 11, 2018 9:02 pm
by orzelek
I fixed all the issues I could find with your debug code in new version.
I have one question - if I use rso-regenerate command from console this debug code will write that regenerateCommand global is used before being defined. But it's a name of defined function so I have no idea why it writes that. Any idea what would be causing this?
Re: Not desync safe!
Posted: Sun Mar 11, 2018 11:56 pm
by Mylon
orzelek wrote:I fixed all the issues I could find with your debug code in new version.
I have one question - if I use rso-regenerate command from console this debug code will write that regenerateCommand global is used before being defined. But it's a name of defined function so I have no idea why it writes that. Any idea what would be causing this?
The debug could should be called as the last thing before the game starts. So at the end of control.lua. Unless you have some strange logic that calls a "require" after the game starts, you might just be setting up your meta table too soon.
Re: Not desync safe!
Posted: Mon Mar 12, 2018 12:49 am
by orzelek
Mylon wrote:orzelek wrote:I fixed all the issues I could find with your debug code in new version.
I have one question - if I use rso-regenerate command from console this debug code will write that regenerateCommand global is used before being defined. But it's a name of defined function so I have no idea why it writes that. Any idea what would be causing this?
The debug could should be called as the last thing before the game starts. So at the end of control.lua. Unless you have some strange logic that calls a "require" after the game starts, you might just be setting up your meta table too soon.
That explains it - I moved it up top. I'll leave it at the bottom (it's commented out anyway since I don't intend to keep it running other then for dev testing).
And thanks for the code - metatables are a bit magic for me that I don't intend to learn for now - I'm more of C++/C# person
