00:03 <russnelson> Hey folks. I want to do a complex animation, where one part moves for time $t=0 to $t=0.2, then it stays still and another part moves for $t=0.2 to $t=0.4, etc.

00:03 <russnelson> Has anybody done something like this before (I expect so), and can you point me to their .scad file?

00:04 <teepee> russnelson: https://openscad.org/advent-calendar-2020/ doors 6 and 24

00:05 <russnelson> My first thought (possibly naive) is to have a time controller module, which splits the time up into chunks, and so when you call this module and give it the sequence number, it returns a time between 0 and 1 .... which stays at 0 until the sequence begins, and then stays at 1 until it ends.

00:05 <russnelson> Thanks, teepee.

00:10 <russnelson> Wow, that code is sublime.

00:57 <InPhase> russnelson: Door 6 was teepee's, door 24 is mine. We both went with the general strategy of calculating a motion segment fraction from 0 to 1 like that. I think it's the simplest way to manage multi-motion approaches, although clearly there are a variety of ways to prepare it and handle it. teepee's anim module with the dynamic $a variable is a particularly clean way to segment it if you can divide

00:57 <InPhase> displaying systematically as in and out. The problem I faced in painting is that half of the behaviors needed to be transient (the different types of brush motion) while half needed to be persistent (the paint).

01:00 <InPhase> Probably you could do something like anim but with negatives instead of undef, and then have it also receive an optional parameter for an activation range, and trigger on that instead of the is_undef approach. That might have been a smarter way to do the painting.

06:10 <RockClimb> Hello

06:13 <RockClimb> I was in a few days ago with a problem with openscad being very sluggish on a multiple video card multiple monitor system when typing. The temporary fix came down moving it to another monitor.

06:14 <ccox_> were you able to figure out why the second monitor was slower?

06:15 <RockClimb> I just wanted to report back that I found the issue  in case you all should see it again. I had one of the smaller monitors set as "primary display" I turned that off it it works great on all 4 now. As for why, I do not know.

06:15 <ccox_> ok. Still sounds like a bug in the video card driver software, though.

06:17 <RockClimb> could very well be. I am running rather old hardware. In my testing I found other apps that were also affected but not as bad as openscad. Turning off the primary monitor setting has fixed all of the issues

06:19 <RockClimb> I think the issue has been there since I switched to manjaro but it was not really noticeable unless you were looking for it..... until I ran openscad

06:20 <gbruno> [github] kintel pushed 5 modifications (Fix crash issue; exceptions thrown when permissions deny access to files opened (#4061)

06:21 <RockClimb> openscad is now 100% responsive and usable on both cards and all 4 monitors

06:21 <gbruno> [github] kintel pushed 2 modifications (Build x86_64 first as that will install compatible helpers, e.g. msgfmt). https://github.com/openscad/openscad/commit/c061f54bb902344c205438914ec95bf5a0055e51

06:22 <RockClimb> I know I will likely forget this fix in a week or 2 but hopefully you younger folks will be able to remember it and be able to help others

06:26 <RockClimb> I am glad to have it fixed though. I was notified yesterday that I have my 2nd part design on the way in a few days. :D

15:42 <peepsalot> teepee: where do the windows dev snapshots come from? is it the circle-ci actions on master?

16:26 <teepee> peepsalot: yep

16:29 <teepee> not quite up-to-date but overview is in https://github.com/openscad/openscad/wiki/CI-Setup

19:17 <veverak> let's say I have one-file model

19:17 <veverak> is there some online viewer I can share it with?

19:17 <veverak> something more than just png

19:17 <teepee> cadhub?

19:17 <othx> cadhub is https://cadhub.xyz/draft/openscad (needs login) or just have it fetch an externally stored script via https://cadhub.xyz/draft/openscad#fetch_text_v1=<URL>

19:19 <veverak> that looks nice

19:19 <veverak> but timeouts on my model :/

19:23 <teepee> I guess that new thingiverse viewer has no direct access

19:28 <teepee> https://models.makewithtech.com/thingiversebrowser

19:46 <buZz> https://gist.github.com/buzztiaan/b48d593f3e3fe9cf96d44bf1bb1e9108 <-- some superuseful hook if you have these kinds of shelvingsystem

19:46 <peepsalot> teepee: it seems like mimalloc didn't override correctly on windows builds, i didn't see any notable performance difference, and can't get the MIMALLOC_VERBOSE environment variable to print stats to console etc.

19:46 <buZz> pretty strong too but i havent measured or put my own weight on it :) its not strong sideways

19:48 <peepsalot> I think I might have to link via DLL, as that is the recommended way to do it for windows, i was just avoiding that since I didn't want to have to deal with install/packaging of additional dll

19:49 <teepee> yeah, normally the build is targetting a fully static exe

19:50 <peepsalot> i'm also not sure if there's anything special that would need to be done for the .com vs .exe setup we have. i know the .com basically just calls out to the .exe but not sure if that interferes with the override methods in any way

19:51 <teepee> the com is really separate and just runs the exe with stdout redirection

19:52 <teepee> so that should not need anything

19:52 <InPhase> veverak: imgur + bpa.st links are pretty reliable while these other viewer options are in intermediate dev states (cadhub) or broken (thingiverse).

19:52 <peepsalot> yeah i just don't understand how the malloc override/redirection works or if that could possibly cause conflicts

19:52 <teepee> I'm not 100% sure, my understanding is the exe is a GUI app so no stdin/out connected normally

19:53 <teepee> where the com is a command line app which has stdin/out available

19:53 <teepee> but that's all some crazy windows magic I don't really want to know much details about

19:54 <veverak> https://pub.veverak.org/kalrack/main.png https://pub.veverak.org/kalrack/main2.png

19:54 <veverak> well, I decided to use pictures

19:54 <veverak> anyway, my rack prototype for kallax is ... well ready to be prototyped

19:57 <InPhase> teepee: I've never needed a .com to get stdin/stdout connected on Windows.

19:57 <InPhase> teepee: Is the problem because we're missing the WIN32 flag on add_executable?

19:57 <teepee> the com is only a trick to have both named openscad I believe

19:58 <teepee> the actual difference is the exe type. I'm not even sure it's *actually* a com file

19:58 <InPhase> Typically one is supposed to do add_executable(OpenSCAD WIN32 ...sources...)

19:59 <InPhase> That tells it to use regular main() as entry, but don't open a console and treat it like a gui app. But stdin/stdout are still connected if it's launched from the command line.

20:02 <InPhase> I've never even once launched OpenSCAD under Windows, but I do Windows builds at work, and it works fine on a direct exe launch with this expected behavior.

20:02 <teepee> $ file openscad.*

20:02 <teepee> openscad.com: PE32+ executable (console) x86-64 (stripped to external PDB), for MS Windows

20:02 <teepee> openscad.exe: PE32+ executable (GUI) x86-64 (stripped to external PDB), for MS Windows

20:03 <InPhase> So then is openscad.com just a wrapper to open a cmd.exe window before launching the gui?

20:03 <teepee> yep

20:03 <InPhase> Ok.

20:04 <InPhase> But the exe will still output to stdout if it is run from an existing terminal?

20:04 <peepsalot> no, the .com will

20:04 <teepee> https://github.com/openscad/openscad/blob/master/winconsole/winconsole.c

20:04 <peepsalot> but if you just call "openscad" from command line, apparently .com is resolved before .exe

20:05 <teepee> yep, that's part of the trick I suppose

20:05 <peepsalot> running the gui exe from terminal immediately returns io to terminal

20:06 <InPhase> But I don't understand why openscad.exe would not already output to stdout if launched from the command line.

20:06 <peepsalot> because windows hates developers

20:06 <teepee> I think that's just how windows gui apps are defined

20:06 <InPhase> But I do this all the time. :)

20:07 <teepee> it's not a GUI app then

20:07 <InPhase> Well it does the gui thing, of opening a gui and no terminal when run by clicking on it.

20:07 <teepee> probably same as java.exe and javaw.exe

20:07 <peepsalot> you have to choose when compiling whether the app uses the window subsystem or console

20:09 <peepsalot> https://docs.microsoft.com/en-us/cpp/build/reference/subsystem-specify-subsystem?view=msvc-170

20:11 <InPhase> I'm aware of the "open a console or don't" thing, but I never encountered failing to route stdout/stderr to an existing console that a program is launched from.

20:12 <InPhase> Am I always doing something different from whatever our build process is for Windows, or are we talking past each other and not discussing the same thing?

20:13 <peepsalot> dunno, got an example app that just works the way you say?

20:16 <InPhase> Well I haven't fired up the Windows vm on this laptop since I switched laptops. Let me see.

20:22 <InPhase> Well... Turns out this vm is in a pretty sorry and badly functional state... That's what I get for trying to avoid home Windows use for years.

20:23 <peepsalot> you running windows updates for the next 8 hours?

20:33 <InPhase> lol. We'll see if it will even do that.

20:34 <InPhase> That would go better if the network connection still worked.

20:39 <Scopeuk> InPhase it looks like under msvc there is a switch to set /subsystem:Console to switch the console behaviour

20:39 <Scopeuk> https://stackoverflow.com/questions/32999694/stop-windows-command-prompt-from-returning-after-an-executable-is-started

20:40 <Scopeuk> https://stackoverflow.com/questions/8030831/windows-application-that-optionally-remains-attached-to-console is probably more detailed

20:41 <Scopeuk> it looks like gcc under msys has some similar switches

20:50 <InPhase> I primarily work under mingw on Windows. Although I've done the same under MSVC, but it has been long enough ago that I wouldn't remember if I had to waste time on a special setting.

20:55 <Scopeuk> more than fair

20:59 <InPhase> Well the libvirt docs (the thing that broke on my VM) have gotten truly abysmal. Paraphrased: "How to set it up: We can verify it is setup with this command. Now, moving on..."

20:59 <InPhase> "A: For extra help see B. B: For extra help see A."

21:41 <JakeSays> what are you guys talking about?

21:48 <teepee> mostly strange windows stuff

22:05 <JakeSays> i spent the better part of 4 decades working with weird windows stuff

22:07 <Scopeuk> this is all rooted in the replacement maloc (mimalloc) that was recently integrated

22:07 <Scopeuk> the replacement function does not overide if directly linked against an exe on a windows application as it does for osx and linux

22:07 <Scopeuk> but it can overide inside a dll

22:08 <Scopeuk> that led on to a potential issue with the windows cli handler (openscad.com) and indirect invocation implications for the dll methos

22:25 <JakeSays> hmm. wonder how chrome does it

22:29 <gbruno> [github] t-paul pushed 1 additions 2 modifications (Add EGL context creation for offscreen framebuffer.). https://github.com/openscad/openscad/commit/434255c9f8828180e9e5374a96aef1c757ae37db

22:32 <ccox_> 2048x2048 surface image in 2021.01: 31 seconds

22:32 <ccox_> 2048x2048 surface image in dev snapshot with mimalloc: 9.7 seconds

22:35 <peepsalot> does the mxe cross-compile count as WIN32 to cmake?

22:37 <teepee> hmm, I'm not sure, it has a separate flag for MXE_CROSS

22:40 <teepee> I'm also a bit confused about: target_compile_definitions(OpenSCAD PRIVATE OPENSCAD_OS="Mac OS X")

22:41 <teepee> ah, it's just used in the NULLGL code