Spring 0.76: pending stuff

[Nemo]

Has anyone tested with .sdd mods? We've been trying to get S44 to work, but it won't sync consistently in the lobby (sometimes it goes between certain people, other times no, no clear pattern between them as of yet). Been using engine rev 5090 and absolutely updated/clean copies of SVN organized S44.

[FLOZi]

To add a little more detail to that:

Yesterday. Me and Spiked do clean installs of Spring 5082. S44 .sdd 667. No sync. Spiked makes a sd7 of those files, and we sync. Peet then joins with his .sdd and syncs with us. During the game I crash out, and Peet desyncs. (S44 667 syncs fine on Spring 75b2 between all of us)

Today. Me and Spiked update our S44 .sdd to 669. We are still on Spring 5082. We sync.

Lurker has a clean install of Spring 5087. We both have S44 .sdd 669. He and I update our exes to Spring SVN 5090. We do not sync. We even try grabbing the very latest unitsync.dll. No sync.

Nemo updates his SVN install to 5090. No sync. He does a clean install of 5090. No sync. He redownloads the s44 .sdd from SVN. He now syncs with lurker, who has made no further changes. I do a full redownload of the S44 .sdd. No sync.


What the hell is going on.

[KDR_11k]

AFAIK sdds don't get checksummed anymore.

[lurker]

AFAIK sdds don't get checksummed anymore.

Then why was I reporting a bad sync with one person and eventually good sync with another? Though I yesterday played rev. 669 with Nemo when it eventually gave a good sync message, today when I forced a fake sync message and tried a game with FLOZi, spring itself rejected the game.

[imbaczek]

I believe that speeding up sdd loading made sdd sync behaviour undefined.

[KDR_11k]

We just had a test game with one unit being disabled through the lobby, as soon as the con that builds it got made the game crashed complaining that it couldn't load the disabled unit.

[AF]

My json unitsync cache program outputs unit data yet for some reason I only managed to output data for FunTA, P.U.R.E and XTA gave no units at all, regardless of whatever combination I put them in.

The program is based on trepans unitsync test program, and as far as I can see it there's no reason within the program why it should show this behaviour. I thought it might be something to do with my loop or not removing all funta references after the first loop, but that doesn't explain why no units are listed at al when funta is removed and the program is reran.

[Kloot]

We just had a test game with one unit being disabled through the lobby, as soon as the con that builds it got made the game crashed complaining that it couldn't load the disabled unit.

Fixed.

[AF]

XTA 9 and PURE 0.5 give luaparser.execute() failed messages when retrieving units in svn. Funta does not do this.

[trepan]

SVN r5098 probably fixes the XTA 9 and PURE 0.5 problem.
The base/springcontent.sdz archive was not always being
included by unitsync.

[KDR_11k]

I think I've seen some weird behaviour from a factory that was stunned while its construction got completed, the constructed unit got teleported back into the factory.

[imbaczek]

I've got a hunch that stutter was caused by r4434. not sure yet, bisecting in progress.

[Auswaschbar]

I've got a hunch that stutter was caused by r4434. not sure yet, bisecting in progress.

I will try to adjust the number of packets sent per second and see if that makes a difference.

[imbaczek]

Just FYI: finished bisecting today:

Code: Select all

$ git bisect bad
Bisecting: 0 revisions left to test after this
[412d855abe9235dcc8853ce6ce616fd26310ccb0] * reduced NETMSG_NEWFRAME spam from c
lients, 3 times / sec are enought for ping calculation * small messages are once
 again send together in one packet

(Note: this is Windows :))

[HAARP]

I noticed that under the latest Ubuntu when Spring crashes 0.75 it somehow doesn't let go of the mouse. I have to hit alt-f2, relaunch spring, start a game, and then quit to get it back.

I'm wondering if this is a Spring issue or an Ubuntu issue.

You may be interested in umouse

[ZellSF]

Uh, AIs setting start positions is nice, but them placing start positions in the enemy box is not :P

Shouldn't it be impossible in the engine to start in the enemy start box? I've noticed this can be done with players leaving with no start position set in the current version too.

Edit: I don't really actually care much about AI play, but should writing .take take all AIs?

[clumsy_culhane]

testing S:44 with Nemo, the SDD still desyncs, fresh installs of both svn spring and S:44 (and we were both up-to-date)

[Tobi]

Think I found and fixed the std::bad_alloc problem. Just doing one last test before commit.

[YokoZar]

Today I had a game go completely nuts because someone was spamming the change speed commands before it started.

By the time the host could issue a lock speed command, messages from me to chat were about 4 minutes behind. While I could see the game happening in real time, it took about 5 minutes of game time before I could start issuing orders again like normal.

Investigating what caused my client to go so bizarre is one thing, but the experience nevertheless reminded me of one important, and simple, request:

Lock speed by default. If only the host's change speed commands worked (until he changed the .setminspeed and .setmaxspeed himself), we'd avoid a lot of problems. It's also much more intuitive this way.

[LordMatt]

Lock speed by default. If only the host's change speed commands worked (until he changed the .setminspeed and .setmaxspeed himself), we'd avoid a lot of problems. It's also much more intuitive this way.

+1 Also show some different icon next to the spect who is drawing peni all over the map so he can be kicked.