On 5/12/10, Per Inge Mathisen wrote: > 2.3.0 has so far seemed like a pretty solid release, and there is no > desperate need for a 2.3.1 yet. Discussions of new features seem to > have moved to trunk. How about we try to schedule a 2.3.1 bug-fix-only > release for mid-June?
We are a bit passed the 'bug-fix only' mantra, since some of the stuff that is currently in svn/2.3 are not really bug fixes per se, but that release schedule sounds OK. > Trunk seems rather stable at the moment, and branching it as a > possible release point before merging any of the big 3 feature > branches may be a good idea. One idea is to branch is to a codename > branch, start merging the feature branches, and depending on how > stable or unstable trunk becomes after that, either proceed to release > from the codename branch, or from trunk after the merges (nuking the > codename branch). While I haven't personally tested trunk in a long time, Zarel mentioned that the pathwalking is broken. Yeah, I know there is a ticket (http://developer.wz2100.net/ticket/1649) but I haven't really had the time to look into this more. The trunk test builds that we threw up haven't really gotten any feedback at all, and with the new web site style, it isn't exactly easy to change things, since we rely on one person, and the information on how to fix/change it, hasn't exactly been forthcoming. > We have 3 big feature branches that we need to merge before > accelerating bitrot makes merging even more difficult. The merge > priority that we have discussed earlier for the big 3 feature branches > is 1) Qt branch, 2) newnet, and 3) lua. Last comes last because it > seems most unfinished yet, and Qt branch first because it has seemed > closest to completion and has the most developers familiar with the > new code. However, we can easily change the merge order - there are no > interdependencies between them. I think we need to clear up some things about these branches. 1) Qt. - we don't really have anyone that is well versed in Qt for the current issues. Those are, on the mac, the input state is swapped from our normal routines. Currently, we can hack around that, however, we still have text issues, and the fact that we can't get key map screen to show Ctrl or Alt like it does on windows / linux for whatever reason. AFAIK, Alt should work like it does on windows / linux, but it doesn't show "Alt +" <letter> like it should. The Main menu / in game menus suffer from a state conflict. That is, when we set modes, (pie_SetRendMode() & pie_SetTranslucencyMode()), it looks like we are conflicting with the way Qt is handling the window/widget. Qt suggest to use beginNativePainting()/endNativePainting(), but that causes a big performance hit, and didn't really fix anything when I tried some tests with that. While experimenting with Qt Designer for a new lobby screen, I don't see a easy way to match the current color scheme / "style" of warzone, and if we go this route, the issue then becomes how easy is it to maintain? I have not done any svg tests, so unsure how smooth this will go. As of right now, I can only think of one advantage to Qt, and that is for the hardware colored cursor support. The lack of a dedicated mac & Qt expert person to help us trying to fix this doesn't make things easy, and it gets annoying, since we must waste more time digging into tons of docs hoping to find out the cause, then waiting for feedback. With SDL, we are pretty much at, "it just works" for all the platforms, so people can concentrate on other issues at hand. 2) newnet. Besides the problems listed in http://developer.wz2100.net/ticket/1786, I don't really understand the main reason for having such a high requirement for precision. (I haven't really looked at the code either) Compiler issues are the main problem here, though, I haven't really played with it enough to see if it has other problems. We can't release test builds of this until that is fixed, and we won't merge to trunk, until we know it works as it should. 3) Lua. The last I recall, the state of lua was about 50-55% done, and the forum thread : http://forums.wz2100.net/viewtopic.php?f=35&t=4733 shows that the conversion script is buggy / broken for some things. We don't really have anyone who is a lua expert either (since Gerard was about the closest thing we had, and he isn't around much), so that is slowing things down drastically as well. 4)Beta widget Yeah, I know this wasn't mentioned, but, it appears there is lots of confusion of the shape this is in. It uses deprecated lib(s), and would add lots more dependencies, and trying to compile those deprecated lib(s) on anything but unix is a huge PITA. It never really got past the (pre) alpha stage, before it fell by the wayside. The thinking to revive this was to use Qt for the SVG routines, but, there have been no tests at all, so we are not even sure if this would be a good move or not. Everything is "in theory, it should work". > If we proceed to make a release from present trunk, we can probably > start the betas for that quite soon and maybe aim for an July/August > release. That depends on no last minute large rewrites of anything > hitting trunk first. > > That means: > June - 2.3.1 (bug fix only) OK... > August - 3.0 (current trunk) We really need to get feedback about the current trunk builds, so, we need to start telling people they can download test builds *now*. Otherwise, we are falling into the beta 200 crap that shot us in the foot for 2.3. > December - 4.0 (qt, newnet, lua, new savegame format) Highly doubtful, without more help. > Any other ideas? > The biggest problem we have is upkeep on the branches. This really needs to be priority #1, since the more we wait, the worse the branches get out of sync. The next biggest problem is, (and I already know there is no easy fix for this), is that everyone still basically does what they want to. We are not pooling all our resources/talent into the same pot. I don't even think this is solvable in a open sourced project. Heck, the other problem is, I don't even know the status on what anyone is doing. One day, a commit will be made, and then you know what that person has been up to. The ML is more or less underutilized, the pattern is, someone says something, then maybe 1 or 2 replies, and that is about it. Ah, the joys of a open sourced project. :) _______________________________________________ Warzone-dev mailing list Warzone-dev@gna.org https://mail.gna.org/listinfo/warzone-dev
