whitequark changed the topic of #solvespace to: SolveSpace--parametric 2d/3d CAD · latest version 2.3 · http://solvespace.com · code at https://github.com/solvespace/solvespace · logs at https://irclog.whitequark.org/solvespace · message channel op if you can't talk
_whitelogger has joined #solvespace
_whitelogger has joined #solvespace
<travis-ci> GU3MS/solvespace#16 (master - b33a635 : GU3MS): The build has errored.
<travis-ci> GU3MS/solvespace#17 (master - c8c4003 : GU3MS): The build has errored.
<travis-ci> GU3MS/solvespace#18 (master - 82218d5 : GU3MS): The build failed.
djota_1 has joined #solvespace
<djota_1> how to get semiminor or semimajor axis of ellipse from eigenvector and eigenvalues?
<whitequark> sorry, no idea, i'm not a math person
<whitequark> i just do c++ and stuff
<cr1901_modern> whitequark: Clearly you've explained this before, but I forgot: Why do you opt to use ANGLE instead of an OpenGL driver for Windows?
<whitequark> cr1901_modern: because if you install fresh win10 it'll come with DX11 on any modern hardware and also OpenGL 1 only
<whitequark> winupdate doesn't carry opengl drivers, only vendor packages do
<whitequark> same reason chromium uses angle, really
<cr1901_modern> Ahhh. Well yea it's been a while since I had to download a vendor package
<whitequark> in theory we could switch to using desktop ogl if we detect it
<whitequark> not sure if that's any better because angle still has to translate shaders
<whitequark> could be slightly less glitchy
<cr1901_modern> why would angle need to translate the shaders if you're using desktop opengl
<whitequark> because gles and desktop gl are incompatible
<whitequark> even desktop gl core profile
djota_1 has left #solvespace ["Leaving"]
<cr1901_modern> ahhh
<whitequark> because gl exists in analternate universe where everything is insane
<cr1901_modern> No argument from me there
<cr1901_modern> In any case, I've found, interestingly enough, for another application, ANGLE is more stable than OpenGL
<whitequark> easily
<whitequark> angle does a lot of validation on its own
<whitequark> because it has to do that for webgl conformance
<whitequark> and it emits fairly conservaive output
<whitequark> i'm probably going to have to enable, like, vulkan in angle
<whitequark> and then on macos enable moltenvk
<whitequark> so it'd be like solvespace->opengl->angle->vulkan->moltenvk->metal->hardware
<whitequark> which is batshit insane
<cr1901_modern> Ahh because Apple decided to stick to it's "my way or GTFO" roots
<cr1901_modern> All*
<whitequark> well, i'm sure they had their reasons
<whitequark> apple uses a ton of standards where everyone else doesn't bother
<whitequark> like zeroconf/bonjour
<whitequark> they even published some of their internal-ish stuff over as RFCs
<cr1901_modern> Interesting
<cr1901_modern> In the past I'd use OpenGL for "portability" and hope for the best. My laptop/main dev machine isn't Vulkan capable (Sandy Bridge), so I haven't really had a chance to learn it
<cr1901_modern> (other than the API looks horrifying)
<whitequark> you're not really supposed to use vk directly
<whitequark> it's like ntdll
<whitequark> well a bit
<whitequark> like syscalls but for gpu.
<whitequark> you want a libc.
<cr1901_modern> Indeed, got any recs offhand :P?
<whitequark> no idea
<whitequark> i don't really care about them i'm not a graphics person either
<whitequark> or a cad person
<whitequark> i just want a usable cad okay
<cr1901_modern> Help, I Just Wanted Usable CAD And I Accidentally Became A CAD Software Maintainer, the Light Novel
<whitequark> story of my life
<travis-ci> GU3MS/solvespace#19 (master - 514e613 : GU3MS): The build is still failing.