OpenMW 0.46.0

Anything related to PR, release planning and any other non-technical idea how to move the project forward should be discussed here.
User avatar
akortunov
Posts: 899
Joined: 13 Mar 2017, 13:49
Location: Samara, Russian Federation

Re: OpenMW 0.46.0

Post by akortunov »

AnyOldName3 wrote: 11 Feb 2020, 10:15 A lot of things there are only true if you make them true. For example, Dolphin Emulator hasn't had a stable release for several years and doesn't intend to have another any time soon.
Actually, Dolphin Emulator is a good example what happens with huge releases - there are too many changes to test them properly, so devs avoid to make stable releases, and users are forced to use nightlies because they have no real alternatives (a latest stable 5.0 release was released in 2016). Technically, it is close to the rolling updates system, which we, IIRC, do not want to use with OpenMW.
AnyOldName3 wrote: 11 Feb 2020, 10:15 Users are expected to use builds from the master branch, and CI artefacts from pull requests are the things for the lab rats. That works out fine while violating a lot of the things you said as fact.
There are build artifacts only for Windows, and I am not sure if anyone really tests every pull request using these builds before it get merged to master.
On other platforms testers have to build pull requests themselves, and I am not sure if testers check every pull request as well.
Who is supposed to check if every of 260 tickets from the 0.46 release works properly and does not cause new issues (to claim the 0.46 to be a bug-free and give us a green light to release the 0.46), and which amount of time such check would require?
From what I can tell, we use the "a game launches, a quick test does not show critical bugs" check before release instead, and it does not guarantee that users will not encounter bugs which were not present in previous release.
User avatar
AnyOldName3
Posts: 2668
Joined: 26 Nov 2015, 03:25

Re: OpenMW 0.46.0

Post by AnyOldName3 »

Dolphin's not just close to a rolling release process. It is a rolling release process. There're the dev builds (not nightlies as they're per-merge, not per-night), which usually get recommended, and there's a monthly update track featuring dev builds that received extra testing for people who are extra conservative.

The 3.0 release was a few months before I started hanging around on their forums, so I can't comment on the process for it. It didn't take long for the recommendation to be to use dev builds (i.e. by the time I joined the forums).

3.5 was partially a trial of a shorter release cycle to see if it made things easier, but was enough of a hassle that it was decided not to do that again. Dev builds were the recommendation within weeks of the release.

4.0 had a feature freeze for a couple of months. Everyone hated it as it killed project momentum, and it still required two hotfixes. It was decided not to do that again, and dev builds were the recommendation within weeks of the release, long before the hotfixes were done.

5.0 was mostly for two reasons. Firstly, it meant that support could be dropped for the bazillion broken 4.0 builds (actually mostly just the 4.0.2 release and one with tweaked netplay for SSBM). Secondly, it meant that Debian would stop giving people 4.0.2. It got a release branch, as the feature freeze for 4.0 was awful. That lasted the best part of a year before it got a feature freeze as the RC was objectively worse than a dev build within a few hours. The recommendation was to use a 5.0-x dev build way before the release even happened.


No one's forced to use a nightly/dev build for Dolphin. They're just genuinely the best choice except for a few hours a year when someone's merged something too early and CI hasn't turned around on the revert commit and you want to play the one specific part of the one specific game that was negatively affected.


We don't have the same huge amount of users, so don't get nearly as much coverage, but I am inclined to think that at least having the forum, IRC and Discord regulars all on nightlies is the right idea. A release build is good for publicity, but it's pretty rarely the best build to actually use, especially before 1.0 when we're adding fixes on the regular.

I guess the point I'm trying to make is that really short release cycles of about a day have advantages over anything longer with the only disadvantage being testing of individual builds, but anything long enough to guarantee good coverage invites the issues of a long release cycle. The only way around that is to live in a hypothetical world where people thoroughly test RCs, but anyone who cares enough to do that would be happier using a nighly. Abandoning releases altogether and switching to a rolling release worked really well for Dolphin, and I think most of the reasons for that are also present with OpenMW.
User avatar
easygamer82
Posts: 13
Joined: 01 Nov 2018, 23:15

Re: OpenMW 0.46.0

Post by easygamer82 »

So....Should I just jump into a Nightly of OpenMW and get on with it or, should I wait for the 0.46 release?

I think the 0.46 release is poised to be a major milestone release for OpenMW, with the shadows and stuff. But that's just my opinion. :D
mikeprichard
Posts: 113
Joined: 16 Dec 2018, 19:42

Re: OpenMW 0.46.0

Post by mikeprichard »

Again as just someone sitting on the sidelines, I don't really see that the official release builds have been significantly more thoroughly tested for each included bugfix and feature than the nightlies, so I'll continue to mostly ignore these releases (other than reading the related news article/watching the video that eventually comes out) and stick to the nightlies. With often a year between official builds, as mentioned above, and no clear advantage (at least in my opinion/experience) to using them over the nightlies in practice, it seems like an easy choice!
onionland
Posts: 68
Joined: 29 Jul 2014, 00:43

Re: OpenMW 0.46.0

Post by onionland »

If openmw is to switch to a rolling release cycle that would be an entirely valid but separate discussion, and should likely also go with a discussion of how other factors such as pr should be changed to reflect that.
The reason Dolphin is able to thrive through rolling releases is in large part to their monthly updates, which keeps the end user up to date with any major progress, and also promotes some media about the new features. Openmw currently does not have that.

Personally I believe a major change to organization should also go hand in hand with a major change in goals, aka, it shouldn't be introduced until after 1.0. The features introduced after 1.0 would also be more suited for more user-oriented active updates.
User avatar
AnyOldName3
Posts: 2668
Joined: 26 Nov 2015, 03:25

Re: OpenMW 0.46.0

Post by AnyOldName3 »

Dolphin was effectively using a rolling release process a decade before they had the monthly auto-update track, and several years before the monthly-ish progress blog posts, so I wouldn't say they're a significant part of why it works for them. Either way, we'd like to have more regular newsposts here than we actually manage, regardless of whether we change our release process.

If OpenMW were to switch to a rolling release process, I think it would only make sense before 1.0. Right now, a lot of development work is fixing bugs rather than adding features (just like Dolphin), and getting those fixes into the hands of the reporter as soon as possible is important for making sure they continue using OpenMW now instead of putting it off for a few months which turn into years. Similarly, if someone comes along and finds a mod they want to use is broken in the current release, but fixed in the next one and available in a nightly if they're keen, they're a lot more likely to not use OpenMW for a year then if they're told it's available in any release from the last couple of months and to just download it. Most changes make the engine more reliable, so breaking people's setups is rare if they stay at the bleeding edge. Once we're post-1.0, it's going to be new features that make up the bulk of things, and as they're not things missing versus the original engine, they're a lot less likely to make people turn away. There's also more risk that any given commit breaks something or doesn't work as intended, so the bleeding edge will be bloodier, and less suitable for general use.
User avatar
raevol
Posts: 3093
Joined: 07 Aug 2011, 01:12
Location: Caldera

Re: OpenMW 0.46.0

Post by raevol »

Sorry if this should be in a different thread, but how would we do rolling release for macOS? Are we currently auto-building those somehow? I see we have nightlies for Windows and the daily PPA for Ubuntu...
User avatar
akortunov
Posts: 899
Joined: 13 Mar 2017, 13:49
Location: Samara, Russian Federation

Re: OpenMW 0.46.0

Post by akortunov »

IIRC, we have nigtlies for MacOS as well.
User avatar
Capostrophic
Posts: 794
Joined: 22 Feb 2016, 20:32

Re: OpenMW 0.46.0

Post by Capostrophic »

User avatar
Ace (SWE)
Posts: 887
Joined: 15 Aug 2011, 14:56

Re: OpenMW 0.46.0

Post by Ace (SWE) »

This reminded me that the Flatpak builds are currently only of the latest release as well, so I'm going to see about setting up the beta branch on them and rolling that as well.
Post Reply