0.2.8.3_beta1: Release process and bugs

[Z-Man]

nemostutlae: duh, yeah, that must be it

The problem is that it relies heavily on new string functionality introduced in the trunk, and we can't really backport that because it would break things all over the place.

That's closer to the truth

[Z-Man]

The system for third party master servers has been added. The general process was simple, it is just a replica of the server bookmark system, just with a different connection function (between pushing the distinguishing information around in static variables and copy-pasting the server bookmark code, I picked the static variables). In the version that is checked in could still be a nasty bug, I had an assertion failure from tArray while not running it in the debugger, but could not reproduce it since then.
I'll put up a master server for competition servers on master1.armagetronad.net:4532; new users will get a default entry for it.

Edit: it's up, and now everyone will get a default entry for it, not just new users. There's already one server using it, guess which

[Rain]

Feature request:

if possible add SCORE_SURVIVE (what you get at the end if you kept alive (personal score))

and FORTRESS_HELD_SCORE (that should be something like FORTRESS_CONQUERED_SCORE but concerning the holding success of the fortress - this will allow to test some interesting scoring systems)


thanks,
kisses

EDIT: and add a "reason" required field in polls posting. thx

[madmax]

Is ping/players/score meant to be able to get colored in the server browser?

0.2.8.2.1:


On 0.3 (7095, the one on AA beta) doesn't happen though:

[Z-Man]

madmax: thanks. I suppose wrtl fixed this bug in the trunk already, or it just doesn't happen there because the server browser works differently. The cause were of course overlong server names with too many color codes. Half a frigging kilobyte of color codes, almost.

Rain: I dunno; what's difference does it make whether you reward survival or punish death (which is already possible)?

About the fortresses, there is no score setting that penalizes loss of one, so there it would make sense in multi-fortress maps. Defeat the enemy, and try to hold as many fortresses yourself as you can. I'll see whether there is a good point to add such a score.

[Rain]

the meaning of SCORE_SURVIVE is to add a personal score for surviving. So if it is a team game, every player that survived takes "n", for example 1 point, finally a team with 4 players that kept alive will get additional 4 points (because the player score goes in team score too). Pratically in a last man standing game, single player mode, it acts just like SCORE_WIN, instead in a team game it gives other scoring possibilities. it is different to death/killing existing scoring because a player that survived could have been more useful in a team work, for example conquering a fortress, than a player that killed 2 enemies, even if that player killed nobody.

FORTRESS_HELD_SCORE could be interesting in time game with multiple fortress. for example a sumo server where a round take 1 minute and, just like a tie, the players who survived get the score.

[Z-Man]

No doubt about fortress_held_score, and you can see it in action on my ladle server right now (it's useful because it is triggered with a delay). And you get the other one just because it was easy once fortress_held_score was there, I still don't see its usefulness.

Edit: Hey, the FORTRESS_HELD_SCORE setting also solves the winzone issue. If a round is ended by taking the winzone, both teams get that score; so with it, you can tune how much a win zone round end will be actually worth.

[Z-Man]

I guess I could use some help on the Mac. Suddenly, the server compilation started to include the client config.h; the only way to stop it from doing so that I found working was removing the client config.h from the project. Of course, then the client does not build any more. It looks like all project headers are virtually copied to a place where they all can be included.

Also, the 10.2.8 build which I hacked produces an app bundle incapable of running; it seems iconv is missing. First building the "regular" client, then the 10.2.8 client gets the missing files installed, but of course I don't know if the result really will be 10.2.8 compatible.

Those two things are just annoyances, they can be worked around.

[Z-Man]

0.2.8_alpha20080209 is up on AABeta and SourceForge.

Tons of stuff are new this time We've got the friends list, redirection kicks, subcultures, and better armathentication support on the client. The Linux and Windows server builds are done with armathentication enabled; since it is still rather experimental, we may still remove it for future builds. For the source version, armathentication has to be explicitly enabled, because it basically has a quirky library source dependence.

I've updated my autopackage tools, and also did one build from Kubuntu Gutsy. That one is the build available as plain GNU/Linux. It should run on all reasonably modern systems and should be your first choice. The old style build is still there and labeled GNU/Linux libc++5, because that's the main difference, you'll need libstdc++ version 5 on your system. It's available on most modern systems as well as an addon package. If you can, please test both versions. I'll be trying to do an universal binary style build next time, maybe ap supports that now.

The mac build was done with the workaround from the last post, and this time, I added the old recorder app I found here on the forums. A domain grabber seems to have gotten dlh's site. It worked for me.

[tonybbb]

awesome...

[dlh]

I guess I could use some help on the Mac. Suddenly, the server compilation started to include the client config.h [...]

I checked in a fix for this.

Also, the 10.2.8 build which I hacked produces an app bundle incapable of running; it seems iconv is missing. First building the "regular" client, then the 10.2.8 client gets the missing files installed, but of course I don't know if the result really will be 10.2.8 compatible.

Well, on trunk I already dropped 10.2.8 support because building new library dependencies is very difficult, and I have no machine to test on. Also, 10.3 was released over 4 years ago... it's time for 10.2 users to upgrade.

Universal ZThread (should work on 10.3.9 also): use attached files to build. I'll put up the compiled binary somewhere soon and update this post. I did run into what I think may be a bug (only on 10.5?). You'll get many warnings over differing visibilities. See Apple's ld release notes (search for "Spurious warning").

[wrtlprnft]

Someone just told me he thinks the alpha sucks because it keeps crashing for them. I wasn't able to get any further details, though :-/

[Loor]

This is pretty neat. How does the registration/global ID work?

[wrtlprnft]

http://wiki.armagetronad.net/index.php? ... or_Players

Matches the real world despide the disclaimer at the top.

Servers supporting armathentication right now:

• Bugfarm Elimination Sumo
• Fortress Café
• ¦×¦.Sumo.Server
• Tag Test

[epsy]

While we're at it, i'ill promote it (again):

You can also use your ¦×¦-Forums username to login.

More info about it here.