0.2.8.0_rc4: Release process and bugs

Website · Post by **Z-Man** » Mon Feb 27, 2006 5:50 pm

Luke: I want to know that, too

The sleep command apparently helps.

Edit: The false player exit message was unique to the branch leading to 0.2.8.1, it was introduced sometime after 0.2.8.0 split from it. It's a network message that should have been delivered to the player you see quitting, but didn't manage to because the connection got deleted. We just throw the messages away now.

Joda's latest revised wall texture looks fine enough for me, I'll use that for the build. The DTD sorting is settled as well. (Philippe has something to say on those, too, but he hasn't been around lately). Tank's crash looks serious, I want it fixed or identified as a weird full moon glitch and a buggy OS before I build.

wrtlprnft · Post by **wrtlprnft** » Mon Feb 27, 2006 7:03 pm

What is this tHeap thing anyways? I got it when I was accessing a nonexistent part of a tString with SubStr(), It took me a while to find out where that error came from... Can't it just crash so I can at least backtrace it with gdb?

Website · Post by **Z-Man** » Mon Feb 27, 2006 7:32 pm

The problem is SDL here, it catches all crashes, shuts itself down (so you're not stuck in fullscreen mode), and exits. The C style exiting then goes wrong, internal datastructures get messed up, and you get that error. If you start the program in gdb in the first place, you'll catch the crash at the spot it happens right away.

Jonathan · Post by **Jonathan** » Mon Feb 27, 2006 11:57 pm

SDL_INIT_NOPARACHUTE perhaps?

Luke-Jr · Post by **Luke-Jr** » Tue Feb 28, 2006 2:30 pm

Are map-0.1[-exp].dtd remaining for a reason? Just not dropping those till 0.2.8.0 is done?
I vote for breaking all maps not conforming to our current specs

joda.bot · Post by **joda.bot** » Tue Feb 28, 2006 2:59 pm

about breaking stuff: I guess it might be better to warn admins and take it down 1-2 weeks later. Also I'm not sure if the master server should hide newer (incompatible) servers, because some old players don't even notice there is a new build out, and they could play race servers etc. ...

This is for a future version though, which supports sorting incompatible servers to the end of the list

.

Luke-Jr · Post by **Luke-Jr** » Tue Feb 28, 2006 8:36 pm

joda.bot wrote:about breaking stuff: I guess it might be better to warn admins and take it down 1-2 weeks later.

Take what down? I'm talking about included resources...

Website · Post by **Z-Man** » Tue Feb 28, 2006 9:23 pm

Jonathan: May be a good thing to activate in debug mode where we don't allow fullscreen anyway.

About the dtd's: The code is still compatible with them (it prints a warning if they're used, though) and some (inofficial) builds can't download them. The same would apply to the beta3 dtd, of course, and it's missing now. To me, compatibility almost is a holy cow: I don't kill it just because I can, only if it is needed. Thus, I'll add the beta3 dtd. In the forseeable future, we can stop this old-dtd-inclusion completely.
Oh yeah: I vote for not breaking old maps just because we can.

Joda: fire up 0.2.7.1 and look at the bottom of the reachable servers (above the long list of unreachable servers)

joda.bot · Post by **joda.bot** » Wed Mar 01, 2006 1:34 am

@luke: sorry, just saw the DTD stuff, I thought you were talking about the (online) resource repository ?

@z-man: was it always there ?

Just heard from some old friends that they could not see the newer servers ?

Luke-Jr · Post by **Luke-Jr** » Wed Mar 01, 2006 3:11 am

z-man wrote:About the dtd's: The code is still compatible with them (it prints a warning if they're used, though) and some (inofficial) builds can't download them. The same would apply to the beta3 dtd, of course, and it's missing now. To me, compatibility almost is a holy cow: I don't kill it just because I can, only if it is needed. Thus, I'll add the beta3 dtd. In the forseeable future, we can stop this old-dtd-inclusion completely.

Even if the DTD is not included, it could still be downloaded.

z-man wrote:Oh yeah: I vote for not breaking old maps just because we can.

It allows us to clear the root resource directory

Also, until the first final release, everyone (should have) knew it was subject to changes, possibly breaking old formats. Once we release a final client supporting these old maps, there may be somewhat of an expectation to keep compatibility with them. But perhaps the warning will suffice here.

joda.bot · Post by **joda.bot** » Wed Mar 01, 2006 10:47 am

Server and Client won't stop working due to local cached copies of the DTDs and maps, right ? So the main task is to get the servers to switch and wait until all old clients/servers hosting games had a fair chance to upgrade.

Players exspect the game to improve upto a release and they don't exspect sudden failures joining a server. (Player's logic "release" is broken.)

1) I think the players switch in time, but not right away.
2) Players might be unable to switch (easily) from say beta4.2.

Are there any beta releases included in distributions ? (I saw some beta4.2 on rpmfind for Red Hat's distributions).

If those break, and there is no working (easy install) update available...

Summary: wait until the release has been out for awhile (IMHO)

Post by **Lucifer** » Wed Mar 01, 2006 11:21 am

OT: I think rpmfind scans sourceforge for rpms somehow, or it has a sourceforge mirror that it scans. It picks up rpms for everything on sourceforge.

Luke-Jr · Post by **Luke-Jr** » Wed Mar 01, 2006 12:04 pm

joda.bot wrote:Server and Client won't stop working due to local cached copies of the DTDs and maps, right ?

Correct, but included resources are not cache. They are included. 'automatic' is cache. Including obsolete DTDs only encourages use, if anything.

joda.bot wrote:Are there any beta releases included in distributions ?

I haven't got the Gentoo ebuilds finished up yet.

Website · Post by **Z-Man** » Wed Mar 01, 2006 3:56 pm

I was hit by Tank's crash as well today testing stuff on my work PC. For me, it's a compiler bug, on GCC 3.3.5 and AMD64 and -Os optimization, every exception throw crashes. I backported the GCC 3.3 detection and CFLAG replacement code used in the trunk, looks like that old GCC really can't handle -Os. Tank: could you verify this fixes it for you?

If it does, that means we're ready and I'll start the build as soon as I have time.

Luke: you did not listen

If a user with "obsolete" libxml builds the game, the DTD's won't be downloaded, so if we want old maps (and OLD here means using the beta3 dtd or earlier) to continue to work there, we have to include the old dtds exactly where the code wants them to be. Faced with the choice of extended compatibility vs. a tidy root resource directory, I choose the compatibility.
And remember the big fat warning when using an obsoleted DTD. I'd say this has to be discouraging enough. We can turn that warning into an error message now on server and standalone games.

I apparently did not make it clear enough that I *want* the default for catecory to be "unsorted". I reverted your change when I fixed the version (whoops).

Website · Post by **Z-Man** » Wed Mar 01, 2006 4:51 pm

Lucifer wrote:I don't mean to post so soon, but it'd be nice if I could do something like:

armagetronad-dedicated --prefix

I have that implemented, but then started to wonder whether this would not be better kept in an armagetronad-dedicated-config script. It sounds like you want to use it like one

Some imaginary guy wrote:XXX-config strips are for libraries that need to publish prefix and compiler flags and linker flags and have no executable that can do this. We have an executable and only need the path to be communicated, a script for this would be overkill.

Since it'll be easier for me to just commit the work than implement that script, I can follow this argumentation. Committed.