0.2.8.2_rc1: Release process and bugs

[Z-Man]

Done so far:
Created the branch b0_2_8_2 leading to this version, the root of the branch is tagged as Root_b0_2_8_2.

alpha20060507 is looking quite good, 0.2.8.2_rc1 will be almost the same code unless something bad happens.

I'll run the usual compatibility tests (various Linuxes, BSDs and Windows) on b0_2_8_2, if they work well, the version will get tagged and built this weekend.

[Z-Man]

I've merged everything up to Root_b0_2_8_2 back into the trunk. I didn't touch the gentoo ebuild, the conflicts there were beyond my understanding. CVS did some interesting things this time, like putting a block from one function into a totally different function in tString.cpp. But in the end, this was less painful than the last merge.

[wrtlprnft]

Ah, thanks, now I get the chatting spectator mode too

[Z-Man]

The build is rolling. I'll give status reports on IRC, pretending someone is interested.

[Z-Man]

Release is on hold. CVS Test crashed With the code that was supposed to go into the release. I think the future wall killing is messing up grid data structures while they're processed.

[Z-Man]

Yep, that suspicion turned out to be correct. It's in fact a very old bug and it wouldn't have been tragic if it had slipped into 0.2.8.2. Of course, now that it's found, it needs to be fixed. Can't make a release with known crashes.

Plan: The next restart of CVS Test will run the fixed version. I'll let it run for two days or so to see if huge new problems are ripped open by this, if not, there's going to be yet another alpha. If yes, I'll make a new plan If no problems appear on CVS Test or with the alpha until the end of next week, I'll restart this release. Objections against recycling the tag v0_2_8_rc1 for this will be noted, but otherwise ignored.

[Z-Man]

The plan seemed to work. With only some minor corrections, the last alpha has turned into _rc1. The corrections in detail:
- A very nasty VisualC error has been worked around. Nevertheless, the VC6 build is not readily available for download. I say we *force* people to try the GCC build and report back errors.
- The default of CYCLE_RUBBER_MINDISTANCE is now 0.05 both in the code and in the config files. The old default value of 0.01 caused technical problems at times, the value in the config files of the last alphas (0.2) makes ironing a too effective deep grinding tactic.
- The autopackage client installer no longer warns you that /etc/init.d is not writable if you run it as a regular user; it isn't trying to install anything there anyway.
- The autopackage uninstaller now correctly cleans up the files installed by sysinstall
- The autopackage script is hacked to default to text mode installation. Downside: installing it from the file browser always fails with a cryptic error message, before the hack, it failed silently most of the time.
- The server installer prints the correct name of the user it creates to run the server as.

The download is available on SourceForge and aabeta. The Freshmeat announcement has been but out.

Tank, could you put it on the download page? Same as with the last RCs, only this time, without the VC6 builds.

[Z-Man]

For those who missed the alphas, here's the complete list of changes since 0.2.8.1:

Code: Select all

- Timestamps and teamscores added to scorelog.txt
- Dedicated server now works on FreeBSD and OpenBSD
- User running the dedicated server is called "armagetronad" again, the
  longer "armagetronad-dedicated" caused problems with BSD
- Test versions refuse to connect to servers more than one version ahead
- Prepared client for respawns
- Team spawn formation is now configurable
- Added reasons to bans
- Added spectator autokicking
- Added history to chat and console (wrtlprnft)
- You only enter a game once your client is synced
- /msg now prefers exact matches over partial matches
- Cycles now have better memory for pending turns (wrtlprnft)
- Added player join/leave/rename messages to ladderlog.txt with IPs
- Ping variance influence on packet loss tolerance code is now clamped
  by the regular, configurable, packet loss tolerance: effect of variance
  can be no bigger than the effect of ping.
- Spectators are now regular players and can chat
- The /msg command now converts the entered pattern into the internal
  name so if you write "/msg öäü" and there is a user named "öäü" with
  an internal nick of "oau" it will still get matched
- Team changes are now only executed when allowed, the need for the 
  temporary/permanent team imbalance setting distinction has gone away
- Added brake toggle button
  Bugfixes:
- Direct cycle collisions sometimes lead to both passing trough each other
- Init scripts now installed correctly with --enable-initscripts
- Hopeless pending team change wishes are deleted
- Large timesteps are now broken down into smaller ones on the game level 
- Pretend keys are released before entering chat
- Two cycle steering events were sent if you doublebound the brake
- Large speed boosts out of game rule bounds were possible during network
  glitches
- Debug recording reliability increased, function call with undefined return
  value eliminated
- Recording and scorekeeping no longer conflict
- Input during a recording now has an end marker, making quick keypresses
  after leaving a menu not break it
- CYCLE_RUBBER_DELAY was overridden sometimes by packet loss tolerance
- ALLOW_CONTROL_DURING_CHAT now ignores modifier keys like shift and
  has been reactivated
- The camera sometimes got stuck in free mode for the first round you watch
- Crashfix in hole blowing code when no cycle was set
- After a "perfect" 180, you're more often on the right side of your own wall
  That code caused some havoc at first, so it has been reworked
- The smart cam did not like fluctuating framerates
- Improved debug recording reliability: multiple master server visits
  and too quick keypresses were causing trouble.
- Canonical DESTDIR support and added ROOTDIR support for testing
- Login floods are detected and ignored
- Player name updates sanitized
- Disabled complicated feasibility tests for team menu entries, 
  they did not have the full information and were often wrong
- Team score only added to player score if no teamplay is possible

As for upgrade recommendations:
Users of the last alpha, 0.2.8_alpha20060515, don't get anything new. They can stick to it if they don't mind the long version number.
Users of the previous alphas should upgrade to rc1 for the bugfixes.
Users of 0.2.8.1are encouraged to upgrade if they want to help with testing.

Plan: If only an ignorable amount of bug reports comes in, this is going to get rebuilt as 0.2.8.1 final next weekend. I think chances for that are good, otherwise this wouldn't be rc1.

[k]

- A very nasty VisualC error has been worked around. Nevertheless, the VC6 build is not readily available for download. I say we *force* people to try the GCC build and report back errors.

Uncle. I tried out the GCC build for a little while this morning and I haven't had any problems with it yet.

[Z-Man]

Heh I'm more worried about total system incompatibilities, actually, "does not run at all" types. Good to see you're still paying attention, though, K

I merged the changes up to this release back to b0_2_8 and the trunk.

[2020]

not sure where to download this version
rc1
for mac ppc
since the beta page has only got
2006-4-something-or-other as an option...

i apologise for my incompetence...

[Z-Man]

There isn't a Mac build yet. Nemostultae, do you have time for that? I don't expect build related problems since the last alpha that got a Mac build, so no problem if you want to wait for 0.2.8.2.

[dlh]

0.2.8.2_rc1.macosx is on sourceforge now.

[2020]

thanks
nemo

it works!

[belenus]

I've found a REALLY nasty bug... I just updated the 0.2.8.1 release to 0.2.8.2_rc1 and during "make install" the ownership of /var/run and /var/log and all contents got transfered to armagetronad.

I was lucky to leave my SSH connection open because I couldnt create a new connection anymore.

Ok problem is... /etc/arma*/rc_config is using NEW PIDDIR and LOGDIR at /var/log and /var/pid, problem is rcd_serverstarter chowns those directories to the user armagetronad.

Solution:
make LOGDIR and PIDDIR a subdirectory of /var/games/armagetronad (uncomment the options in rc_config)

I really hope that chowning whole /var/run and /var/log back to root doesn't get me new problems now with other programs... *grrrrr*

I would love some more testing of stuff like that before you put it up for download, especially when changing paths.