The topics from previous years are closed as new ones are opened. But the current one for 2026 is open to everyone ![]()
Released Drag[en]gine 1.31 together with DEMoCap 0.7 . The HaikuPorts package build also run through after testing a workaround for the build server timeout problem. Package is ādragengine-1.31-2ā. This release is less interesting for Haiku users since it revolves mostly around VR improvements and fixes and Haiku does not have VR support so far. The next one should be more interesting. Depends on how my time schedule pans out.
@Begasus Iām trying to update the haikuports receipt to version 1.32 but Iām stuck. I added these requirements:
BUILD_REQUIRES="
ā¦
devel:libfftw3$secondaryArchSuffix
devel:libopus$secondaryArchSuffix
ā
But test building fails with error:
Reason: build-requires ādevel:libfftw3ā of package ādragengine-1.32ā could not be resolved
I verified against the fftw receipt and Iām using the exact name listed in the receipt. Any ideas why this is not working?
@dragon Iām not seeing anything wrong with it, even added both in current recipe (I know itās not used there) just to be sure itās not a pkgman issue, could you post the wip recipe you are using somewhere to check it out?
Build launched fine here?
Using that repository and running haikuporter gives me this error. I have installed the fftw and upos package so I donāt know whatās wrong. Technically the build works if done manually in my code repository but the test with haikuporter gets me nowhere right now.
Not sure whatās going on on your end then, build completed fine earlier here.
Could you paste the build log you have there?
Most probably haikuporter somehow got confused over here. Uninstalling and installing the libs a couple of times seems to have unblocked it.
Released Drag[en]gine 1.32. For Haiku the added audio analyzer (lip sync) and stencil shadows applies. The Wayland and HDR monitor support is not present obviously.
And thatās what Haiku Lip-Sync would look like (MESA pipe under VirtualBox):
Nice cool video!
![]()
Released Drag[en]gine 1.33 and DEMoCap 0.9. For Haiku this means in general stability fixes, improved shadow casting and reflections besides other fixes. The rest is about the editor (which does not yet run on Haiku) and the VR MoCap tool (no Monado on Haiku either). But still worth testing with that nVidia driver (if you can get the driver working).
It seems there are problems with the Haikuports build server using new rules. So for the time being it is required to build manually until this is fixed, which hopefully does not take long.
The build server build works now again. 1.33 hpkg is now available.
Released Drag[en]gine 1.34 : Drag[en]gine 1.34 and DEMoCap 1.10 released ā DragonDreams . Mostly VR stuff not relevant for Haiku but AVIF file format module has been added for higher compression needs.
Copilot is listed as a contributor on the GitHub project, but the contributions graph doesnāt show a graph for Copilot contributions.
Is this just GitHub doing the thing where they show Copilot as a contributor on just any old project whether it has commits or not?
There have been test-branches where I tried things with copilot agentic, in particular my initial refactoring test to see what copilot agentic can do and what it can not do. But in case this is your concern, no code is committed to the productive branch that I have not written or reviewed myself, and this includes all kinds of sources like stack overflow, example code and so forth.
Thatās a load-bearing āorā there. ![]()
(I personally avoid LLM-generated code completely, with as close to a āone drop is too manyā level as I can get; human review doesnāt change it in my opinion.)
But the forum policy on which category the software belongs in is quite a bit looser as to how much means it belongs in āOtherā; direct commits by an LLM-driven account like Copilot would certainly do it, but substantive human-reviewed contributions being LLM-generated would also.
Iām not interested in policing your methodology, just in having the thread be in the correct category, so the best choice is to leave it up to you to say whether it contains LLM-generated code.
(It does look neat; if I hadnāt just started on Godot with my kids Iād be tempted.)
This game engine has been written by me over the years. I have written the 900k+ lines of code. Copilot inline completion, bug hunting using copilot (especially on Windows with a total train wreck of debug facilities) or other IDE features of whatever kind doesnāt change this. So no, this is not AI generated slop software and does not belong in that category.
(and it is never too late to look at other engines⦠some projects switched engines 3 or 4 times
)