openmw.org

The most important question, is OpenMW ready for OSG-3.6?

I've been working behind the scenes with other Debian Devs on packages that OpenMW, one in particular is OSG-3.6.

It is almost ready for being introduced into Debian and picked up by Ubuntu for their next release (the one after Cosmic).

My latest success was to un-break OSG on armel/armhf (Raspberry Pi):
https://salsa.debian.org/openscenegraph ... requests/1

I'm hoping to get it backported to OSG-3.4 but that there might be more push-back there.

I have been successful in getting many of the patches from scrawl and others that never made it upstream, because of Robert, into Debian's version of OSG.

There hasn't been much of any work done lately on OSG side, but I just want to make everyone aware... if there is a bug in OSG and you think things can be patched. Feel free to do so in our fork and we can test it out. Feel free to also reach out to OSG if you wish to upstream. However there has been at least 2 instances where Robert has been belligerent against AnyOldName3 because of his nick. As a work-around, might I suggest we use MrOpenMW as our point of contact? Perhaps that might be better? I doubt we can change Robert's behaviour, but in order to get shit done, we can at least present any questions, PRs, code as from the OpenMW project as a whole? That might carry over better.

Currently OpenMW crashes with OSG 3.6.3 (at least, on Linux). I suppose this issue needs to be addressed first.

Maybe, if 3.6 branch of OpenMW/osg could be rebased with all the patches made previously plus some new 3.4 branch patches to the release branch of 3.6-whatever. But it's like asking whether OpenMW is ready for C++20C++17.

That's why I'm asking, I remember reading through some merge requests awhile back talking about crashes. I believe it was this one:
https://github.com/OpenMW/openmw/pull/1715

I think we should try to get OpenMW working with upstream 3.6.2 (latest stable release) first.

The point of our fork was for performance reasons, but if there is an bug that causes OpenMW to crash in upstream, then it needs to be addressed upstreamed and not just in our fork.

Update: There is no push or urgency behind this, OpenMW works fine with 3.4 and that isn't going anywhere for the time being.

If we can get it to work, I'd be in favour of OSG 3.6. Switching from display lists and immediate mode to VAOs should mean that we're using parts of the GPU drivers which actually receive attention, so might bring a bunch of benefits.

I remember I did do some preliminary work getting the OpenMW OSG branch updated to OSG 3.6.3, which is where I initially found about the regression from that one commit. I think I still have the work that I was doing on it, and I think I did put in a pull request at one point against the fork before discovering the crash and taking it down while I tried to find the source. It probably would be for the best to try to get that regression fixed upstream, as not all distros and the like will be having a choice of our fork.

Getting the patch into distros helps to reinforce the importance of the fix to upstream. I can help there since I help the maintainers of OSG, OSG-3.4 and shortly OSG-3.6. That means that the patches then get replicated across Ubuntu, Mint, Raspbian and more. It is usually only someone really obstinate that would refuse to fix an issue so widely patched.

For those interested, I've got OSG-3.6 (3.6.2) in staging:
https://launchpad.net/~openmw/+archive/ubuntu/staging