cant sync some mods on LAN in 076b1

[Smiff]

first congrats and thanks on 076b1 release.

unfortunately, there's some nasty network code problem that's stopping us playing several good mods, on the LAN, on the new version of Spring (mods which worked ok in 75b2). the problem has been in every SVN version also since i started testing that, about v4999. sorry about starting another thread but maybe something specific will be seen by other players with the same problem.

mods affected include:
CA
Evolution RTS
XTA

mods not affected (always start fine on our LAN) include:
Simbase
PURE
BA?

mods uncertain:
Spring 1944 (sometimes takes 2 tries to start, although this mod is generally unstable - i think there are other problems)

i don't see understand we should be having networking problems with two modern-ish PCs on a gigabit LAN..?

PC1 (Server): dual core athlon 4200+, XPsp2, 1gb RAM
PC2 (client): sempron 3000+, XPsp2, 1gb RAM

symptom:
server starts game
server reaches "finalising" very quickly
client reaches finalising (some time later, maybe 30 seconds)
server reports client timed out.
client starts as if everything is ok.
server crashes or quits.

this happens repeatedly, although sometimes starting a 2nd time after the crash gets us going (sometimes works with 1944, never works with CA for example).

does anyone else see this problem or have any ideas please? at first i thought it was due to the much faster loading on the server, although i'm not sure.. perhaps two PCs with similar Spring loaing speeds wouldn't see this problem? also i don't know why Spring loads so slow on that client PC, as it actually has a faster newer hard drive (7200rpm SATA2 Samsung, 16MB cache.)

edit2: XTA also affected :/

[imbaczek]

Much faster loading of the server is indeed the issue.

Just a thought: does the client have an intrusive antivirus (like norton?)

[Peet]

mods not affected (always start fine on our LAN) include:
Simbase

You're set then

[Tobi]

Hmm strange, I managed to reproduce & fix that bug, same for Auswaschbar who actually was somewhat faster with committing :), but apparently we fixed the wrong bug after all.

[Smiff]

thanks for all your attention guys.

what's really frustrating is seeing the client sit there on finalizing, knowing it's going to time out, and wondering - what is it doing in there?!
maybe, you could timestamp the logs and add some more details please because my other question is why is it so slow to load some part on some systems.

[Forboding Angel]

Finalizing is when the lua is being initialized.

Evolution RTS does use a significant amount of lua, but it pales in comparison to the amount of lua in CA.

[imbaczek]

It does what the mod tells it and that's a problem. It may do a lot of calculations or a lot of disk reading (or both) - that's why I was asking about an AV, some can slow disk access down by as much as 3000% (that's three thousand percent.)

[aegis]

some can slow disk access down by as much as 3000% (that's three thousand percent.)

use nod32 for great justice

[Smiff]

yer i have AVG on both PCs, it's not that. there is something wrong with that PC though that i've never figured out. the HDD benchmarks over 84MB/s - (which is a faster drive than in the server, which loads Spring really quick) yet it thrashes (pages) like crazy in real apps ... it's a mystery. anyway:

i guess the simplest fix might be to make the server wait much longer for slow clients?

(unfortunately XTA also seems to be in the affected group :/
i am going to try a clean install of 076 later and reboot that machine...)

[Smiff]

did some more testing.

if i use a different PC as the slower client, it still doesn't start properly. so it's not a specific problem with one very slow PC.

but, if i use two slower (not really slow, both 2Ghz ish with fast drives) PCs as server and client, then Spring still doesn't start CA properly! so this seems to be more serious actually than I thought.

my conclusion is that it's possible for something to take so long in Finalizing that it will fail to connect any network clients, regardless of how fast those clients are?

i also tried clean installing 76b1 and that made no difference either :/

[aegis]

have you tried hosting with the other computer?

[Smiff]

ok, right.

the only way (?) i can get CA to work, is to run 076b1 server on a slow pc, and the client on a much faster one. then the client has started first and everything seems ok (after tons of GameID msgs, is this normal?)

Code: Select all

Client connected on slot 1
Player slow_pc_server joined as 0
Player fast_pc_client joined as 1
GameID: 84437c4766911093897bea87d46b2aa7
GameID: 84437c4766911093897bea87c69d49b0
GameID: 84437c4766911093897bea87c69d49b0
GameID: 84437c4766911093897bea87c69d49b0
GameID: 84437c4766911093897bea87c69d49b0
GameID: 84437c4766911093897bea87c69d49b0
GameID: 84437c4766911093897bea87c69d49b0
GameID: 84437c4766911093897bea87c69d49b0
GameID: 84437c4766911093897bea87c69d49b0
Max Windmill altitude bonus is: 22%

edit: to be clear, this log is from the fast PC which was running as client this time.

other combos like running both on slower PCs doesn't work.

of course, serving on your slow PC is not normally done :p but at least anyone can run bots in Spring.

edit: and it's just dawned on me, that the reason this PC is much faster is down to CPU/RAM speed while initing Lua, not hard drive/loading files.

[LordMatt]

Tonnes of GameID messages can't be good.

[Auswaschbar]

Sure you receive lots of gameID messages because the server resents them because he didn't get an ack from the client.

The main problem is the client doesn't update his network while loading and thus not sending any keep alive packets. Strange is that the update runs in a different thread while loading and I have no idea why the updatethread never runs.

PS.: network timeout is set to 30s

[Tobi]

It may be doing the stupid frameNumber == 0 let's do another huge amount of loading? (When the update thread has already stopped)

[Smiff]

can i help any further? please let me know.

edit: great, i see you are doing spring_0.76b1+ builds. i will continue testing them and report any changes, although I might not have time to test every day sorry.

are there any other differences between release builds and SVN builds?
edit2: ok, SVN versions can't use the public server. that makes sense, but i'm looking forward to the bugfix release!?

do you need any more crash logs or anything? or is this clear enough. thanks!