<JordanBrown[m]>
kintel: teepee: this is a previously-working repo, and the file doing the include is submodules/mimalloc/src/stats.c .
J23879218 has joined #openscad
J238792 has quit [Ping timeout: 260 seconds]
Guest69 has joined #openscad
LordOfBikes has joined #openscad
<JordanBrown[m]>
kintel: teepee well, I *thought* it was a previously-working repo. Looks like the problem was that the directory name has a comma in it.
<JordanBrown[m]>
I swear that's the directory name I've been using for this project for a while.
<JordanBrown[m]>
Now if only I knew how to set that flag :-)
<JordanBrown[m]>
-D...=ON?
<peepsalot>
yeah pass it during the configure step: cmake -DGLOBAL_DEPENDS_DEBUG_MODE=ON ..
<peepsalot>
should work i think
<kintel_>
Looks like something happened to boost though: old: Boost: 108100, new: Boost: 1.81.0
<kintel_>
Apart from that: OpenCSG: 1.4.2->1.5.1 Qt: 5.15.7->5.15.8
<JordanBrown[m]>
CMake Warning:
<JordanBrown[m]>
GLOBAL_DEPENDS_DEBUG_MODE
<JordanBrown[m]>
Manually-specified variables were not used by the project:
<JordanBrown[m]>
:-(
<kintel_>
If you can find the list of libraries being linked into openscad.exe, it's usually possible to deduce who dragged them in based on the position of the lib
<kintel_>
Unfortunately, the logged cmd-line contains stuff like @D:\a\_temp\msys64\tmp\ccXRcEmI so I cannot see it from here
<kintel_>
..but it could be quicker to assume boost is the culprit and look at D:/a/_temp/msys64/ucrt64/lib/cmake/Boost-1.81.0/BoostConfig.cmake
<JordanBrown[m]>
That doesn't mention ICU.
<JordanBrown[m]>
I think the change is that the ICU library configuration changed.
<kintel_>
The icu package itself didn't change though
kintel_ is now known as kintel
<JordanBrown[m]>
if you google libicui18n and libicudata, there are lots of people saying "where is it?".
califax has quit [Remote host closed the connection]
<kintel>
yeah, so two possible problems: Some package is linking libicu wrongly, or libicu should have been updated in msys2 but wasn't
califax has joined #openscad
<JordanBrown[m]>
yep
<kintel>
Here's a list of all differences in msys install between a working build and a broken build: https://bpa.st/BX7KW
aiyion has quit [Ping timeout: 255 seconds]
<JordanBrown[m]>
and no changes in OpenSCAD between the two?
<kintel>
no, same git hash
<kintel>
I simply reran a previously working build and it broke
<kintel>
My bet is that the boost-generated CMake config is broken, or is missing to take into account the actual dependencies needed
<JordanBrown[m]>
yeah, but so far I haven't found anything other than ordering in the command line to tie Boost to the "-licui18n" and "-licudata" in the link command line.
<kintel>
Did you find the boost CMake Config file?
<JordanBrown[m]>
Huh. I don't think I've found that one.
<kintel>
Looks like boost is making some (wrong) assumptions about how libicu was built
<JordanBrown[m]>
ah, yes, that's among the ones I just found, just with the line broken differently.
<kintel>
yeah, I removed the newline :)
<kintel>
Anyway, that's likely the root cause
<JordanBrown[m]>
Maybe when I did my big greps through /mingw64 I assumed that the consumer was well-behaved and was mentioning "ICU", so as to tie into the ICU support...
<JordanBrown[m]>
but yes, I agree.
<JordanBrown[m]>
we don't know exactly what changed.
<kintel>
boost-regex is the only lib we use that's influenced by this
<kintel>
Basically, in boost-1.81.0-1, the cmake files weren't shipped in MSYS2
<kintel>
..but in boost-1.81.0-2 they are
<kintel>
Easy fix for msys is to roll back that change and try to fix it. I can open a ticket for that.
<kintel>
Not sure how easy it is for us to work around such issues; force-install an old package perhaps, or find an msys2 package which includes the missing libs?
<JordanBrown[m]>
I looked for a while but did not find one.
<JordanBrown[m]>
I just deleted the -licui18n and -licudata from the build.make file, and it built OK.
<JordanBrown[m]>
so they seem to be a bogus dependency, at least for us.
<JordanBrown[m]>
I edited the references out of the boost-regex cmake file, and they don't appear in the resulting build.make.
<kintel>
Cool, so a quick-fix could be to patch those files in msys-install-dependencies.sh if we detect the bad boost version.
<kintel>
..unless they fix this quickly
<kintel>
Thanks for testing, at least we now know what's wrong and how to make it work :)
<JordanBrown[m]>
sure, always happy to help
<JordanBrown[m]>
I feel a little dumb that I didn't find it earlier; once I found the ICU stuff for cmake and how *it* could pull those in, I assumed that the other components were well-behaved and did more restrictive greps...
<JordanBrown[m]>
Or something like that.
<kintel>
I always say that managing build systems is so much harder than writing the software that's being built :/
<JordanBrown[m]>
It can be, indeed.
<JordanBrown[m]>
at work (Oracle Solaris), we have several senior people dedicated to maintaining the build environment.
<kintel>
heh, I wonder how much bribe they receive for not rage quitting every monday :)
<JordanBrown[m]>
Indeed.
<JordanBrown[m]>
But we do have the advantage that much of the system is entirely under our control.
guso78 has joined #openscad
<guso78>
Hi, I am the crazy guy with the python branch.
kintel has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<JordanBrown[m]>
"the python branch"?
<JordanBrown[m]>
the evalstring thread?
<JordanBrown[m]>
Ah, read a little deeper.
<JordanBrown[m]>
Yeah, an interesting idea. I did something a while back, with a Python library that generated xcsg files that could then be processed by Carsten Arnholm's xcsg. It worked, but Python has a huge problem: you can't sandbox it.
<JordanBrown[m]>
You can take somebody else's OpenSCAD program and run it, and be pretty confident that the only bad thing it might do to your system is to hang OpenSCAD.
<JordanBrown[m]>
Not so with a Python program; it can do arbitrarily bad things limited only by the OS restrictions on the user running it.
<guso78>
Hi Jordan, nice to meet you here. You are from America ?
<JordanBrown[m]>
Yes.
<JordanBrown[m]>
Los Angteles.
<JordanBrown[m]>
Los Angeles
<guso78>
haha, i was once there when i was youth ... from LA to San francisco haha
<JordanBrown[m]>
Bedtime soon, as soon as my OpenSCAD build finishes.
<guso78>
suppose its 11:30pm in your place
<JordanBrown[m]>
yes
<guso78>
right now pythifying scale and mirror function :)
<JordanBrown[m]>
translate, rotate, scale, mirror should all be almost the same; the only difference is in the transformation matrix.
<guso78>
what i like most about the python in openscad that you could e.g. calculate the pixels of a qr code on the fly and dont need to create qr code strls in 2 stewps
<guso78>
yes i know, this task is easy
<JordanBrown[m]>
I don't know QR code encoding, but I expect that you could calculate it in OpenSCAD.
<guso78>
The algorithm behind qr code is quite huge. i am sure you can do it in openscad if you spend enough code and effort. but there are ready-to-use functions in python libs
<JordanBrown[m]>
Library availability is certainly important.
<guso78>
you mean a powerful addition to openscad capabilities
<JordanBrown[m]>
Oh, I think that a Python- or JavaScript- based CAD language would be a big improvement in several ways.
<JordanBrown[m]>
Python has the sandboxing problem.
<JordanBrown[m]>
They both have the problem that the minimum program is bigger than I would like.
<JordanBrown[m]>
The minimum OpenSCAD program is "cube();", and in the other languages would need to be more.
<guso78>
i see. is this what you refer to "sandboxing problem" ?
<JordanBrown[m]>
No, the sandboxing problem is restricting what the program can do, so that it isn't a security problem.
<JordanBrown[m]>
You can't (or shouldn't!) run a random Python program that you download, without inspecting it carefully, because it might be malicious.
<JordanBrown[m]>
You *can* run a random OpenSCAD program, because even if it's malicious it can't *do* anything.
<guso78>
i see the issue. let me think about that
<JordanBrown[m]>
People have tried to sandbox Python, but have not succeeded; there have always been leaks.
<JordanBrown[m]>
That's not to say that a Python-based solution isn't valuable, but you can't just grab models from someplace like Thingiverse and run them.
<guso78>
security of openscad is not important to me personally because i write the code myself.
<guso78>
openscad could issue a warning when the code includes any import statement. there are still soo many powerful things that python can do without importing libs
<JordanBrown[m]>
I believe that Python programs can do bad things without ever doing imports, but I'm not sure.
<JordanBrown[m]>
And yes, for many use cases it's not a problem.
<guso78>
without import, it cannot open ports, cannot browse the internet, cannot access the filesystem, but i am not sure either :)
<guso78>
and if it accessed the filessystem it would only run stuff, which is already there in the computer
<JordanBrown[m]>
You might think. But people have found ways to do exciting things using the introspection mechanisms.
<JordanBrown[m]>
I don't understand the details.
<JordanBrown[m]>
running stuff that is already on the computer can be interesting.
<guso78>
hmm, give me some time to google that ...
<JordanBrown[m]>
running "rm -rf /" makes your life interesting.
<guso78>
yes, but a user does not have permission to do that, and linux user normally is aware of privilege separation
<JordanBrown[m]>
Try it. You might be surprised how much havoc an unprivileged user can cause.
<JordanBrown[m]>
In particular, you can destroy your own $HOME.
<JordanBrown[m]>
But in a multi-user context, how many directories do you share with other users?
<guso78>
yep, got it. i will keep the sandboxing problem in the background
<guso78>
i have done git pull, then compiled my master, evalstring still works as expected, then git push my master back to gsohler/openscad
<guso78>
did not see a "synchronize" button anywhere. if you expect something else, please detail :)
Meido has joined #openscad
<Meido>
https://pastebin.com/07qwzwZa So when I run this code the first shape in the module renders and is affected by the difference but the shape after is generated as a positive image...i'm going to be using this exact pair of shapes SEVERAL times so if there's a way to make them work in a module it would make my code far cleaner...is there a trick i'm
<Meido>
missing?
<guso78>
windows build for evalstring failed because its missing lib -licudata. was there a new lib dependency introduced by me inadvertently ?
<Scopeuk>
guso78 the ci system started playing up yesterday, I don't know how far they got to resolving it overnight
<Scopeuk>
it was definately licudata it was complaining about then
<Meido>
gahh. this 'glitch' is giving me a headache...
<Meido>
if i put the two cylinders in the main body of the code inside of the difference they work as expected...but trying to combine them into a module just causes it to fail consistantly i feel like there must be a way to make this work or what is the good of modules?
<Scopeuk>
Meido in what way fail?
<Meido>
so the second cylinder seems to ignore the difference(){} call even though everything in the module should theoretically be inside of that call.
<Meido>
instead of being cut out of the cube and effectively becoming a negative shap like the first cylinder it is rendered as a positive shape effectively removing some of the other cylinder in teh process.
foul_owl has quit [Ping timeout: 260 seconds]
<Scopeuk>
ok, could you paste bin your code somewhere?
<Meido>
of note the union and difference in the module were me trying to see if either of those would force the two shapes to act as one.
<Meido>
it behaves the same with them as weithout them.
<Scopeuk>
ok that union will prevent the differnce doing anything
<Scopeuk>
the results on the union is a single thing
<Scopeuk>
so your current difference is differencing that combined shape with nothing
<Meido>
ok like i said even without that difference the code still doesn't work.
<Scopeuk>
to think through what is supposed to be happening MagHole should be the "hat" shape and then you want to take two of those out of the main body?
<Meido>
yep
<Meido>
hat shape i like that. it is kind of top hat like innit?
<Scopeuk>
so without the difference it appears to work for me (in the rendered stl) just not very visible inside openscad
<Meido>
i added the union thinking that would make the two shapes one like you said but even with it for some reason only the first cylinder gets subtracted from the cube while the other renders solid effectively cutting the wrong shape.
<Meido>
for me the render with union alone the hole is still to shallow for the second cylinder to be included...
<Meido>
since the two cylinders are overlapping it's actually really easy to see when the second one is being missed by commenting it out and noting how much deeper the hole gets.
<Scopeuk>
maybe comment out the colour, looks ok to me without it
<Meido>
hrm it's still the shallow version...
<Scopeuk>
the shape looks like the result of magHole to me
<Scopeuk>
how deep are oyu expecting
<Meido>
like i said if you comment out the second cylinder the depth of the hole doubles.
guso78 has quit [Quit: Client closed]
foul_owl has joined #openscad
<Meido>
oh wow... i'm so stupid...perpective has been messing with me this hole time...the floor actually looked like it was going up when it wasn't.
<Meido>
thanx for the help Scopeuk and sorry for wasting your time.
<Scopeuk>
you can measure distances between surfaces etc
<Meido>
i generally use sketchup for that. this time i was relying on my eyes too much...i'm a fool for not realizing the vidual cues other programs give me were missing so the floor seemed closer than it was...
<Scopeuk>
it's easily done, particularly depending on what angle you are looking at
<Scopeuk>
I guess "wireframe" might have made it more obvious
<Meido>
is there a wireframe view?
<Meido>
... i found it the other day and totally forgot
<Scopeuk>
view -> wireframe
<Meido>
yeah somehow i forgot that entirely.
<Scopeuk>
also worth knowing you can make any component the only thing visible by putting ! infront of it
<Scopeuk>
I added that to line 22 to see the mag hole in position and then change the design
little_blossom has quit [Quit: little_blossom]
<Meido>
oh that's very nifty
<Meido>
thanks for that tip it'll be very very useful.
<Meido>
that will be some reading for sure. thanx again. something that i've been wondering... when using difference to prevent rendering issues is extendign eh subtracted volume outside of the base volume the best solution or is there a trick where i can have the two volumes flush without rendering with that weird semi tranparent film?
guso78 has joined #openscad
<Scopeuk>
no, its a bit of a maths edge case (when they are flush you technically defined a 3d shape with a zero thickness that remains), extending beyond the edge is the best solution
<Meido>
cool thanx for verifying.
snarkyartichoke has joined #openscad
ur5us has joined #openscad
snarkyartichoke1 has joined #openscad
snarkyartichoke1 has quit [Client Quit]
snarkyartichoke has quit [Quit: Leaving]
snarkyartichoke has joined #openscad
guso78 has quit [Quit: Client closed]
ur5us has quit [Ping timeout: 252 seconds]
guso78 has joined #openscad
J23879218 has quit [Quit: Client closed]
J23879218 has joined #openscad
guso78 has quit [Client Quit]
Meido has quit [Quit: Client closed]
califax has quit [Remote host closed the connection]
califax has joined #openscad
snaked has quit [Quit: Leaving]
<InPhase>
peepsalot: Well there was, in fact, a pretty decent chance I would have tried a ClosePoints version. :)
guso78 has joined #openscad
califax has quit [Remote host closed the connection]
califax has joined #openscad
<teepee>
oh, nice, the boost issue is already fixed by rolling back the patch as kintel suggested
califax has quit [Remote host closed the connection]
epony has quit [Quit: QUIT]
califax has joined #openscad
teepee_ has joined #openscad
teepee has quit [Ping timeout: 255 seconds]
teepee_ is now known as teepee
epony has joined #openscad
foul_owl has quit [Ping timeout: 260 seconds]
foul_owl has joined #openscad
aiyion has quit [Ping timeout: 255 seconds]
aiyion has joined #openscad
linext has quit [Quit: Leaving]
snarkyartichoke has quit [Quit: snarkyartichoke]
guso7810 has joined #openscad
guso7810 has quit [Client Quit]
epony has quit [Remote host closed the connection]
epony has joined #openscad
epony has quit [Remote host closed the connection]
epony has joined #openscad
Guest17 has joined #openscad
Guest17 has quit [Client Quit]
guso78 has quit [Quit: Client closed]
Lagopus has quit [Ping timeout: 260 seconds]
guso78 has joined #openscad
<guso78>
hi together
<teepee>
hi
<guso78>
user or developper ?
<teepee>
like I commented on github, the CI fails are not caused by your change :)
<teepee>
both
<guso78>
haha, you are perfectly on sync
<guso78>
today ported more functions. i think my regular used function set is there now. now offering some of the functions as methods too
<guso78>
wanted to allocate some time to get the CI workflow working in my place. its interesting beeing able to offer programs in "foreign" operating systems , too
<teepee>
and that was my point, as you could just see with the current problems with both windows and macos, it needs to get constant attention
<guso78>
message taken. my 1st priority is getting it working. i believe that keep it working might be easier . haha
<teepee>
nope, considering I'm doing most of the release stuff for 10 years, I can safely say it is not
<teepee>
it's just not so visible to most people as it's boring work that's pretty much invisible to users
<guso78>
ahh, i realize you are the creator of all the great exe files, which i have been downloading for 4 years now .. nice to meet you.
<teepee>
if that's the official releases, that's likely :)
<teepee>
there's still lots of other people working for various other distros / packaging places
<guso78>
one guy each for each linux dist ... but for windows there are at most 32bit and 64bit
<teepee>
well, there's vcpkg, conan, winpkg or whatever all those things are called that I've never used
<guso78>
haha,me neither ...
<teepee>
strictly for release, it's only those 2 builds, but we do have the additional msys2 build for running the test suite and someone maintains a native VS build that we currently have not integrated
<guso78>
suppose the only advantage of the msys2 build is having an unix environment available together with the exe file. shipping msys2 exe files is not very user friendly
<teepee>
it's currently the only way to run the tests in an actual windows env
<teepee>
I had an AppVeyor build earlier that at some point did not make it in the allowed 1h runtime
<guso78>
tests are run by script and they perform all sorts of geometrical stuff ?
<teepee>
yes, cmake has an integrated ctest as actual runner but it basically runs some python scripts that do stuff and compare either the stl/svg/3mf or whatever output
<teepee>
or they generate images and compare those against a reference
<guso78>
suppose there is no way to perform the same in Windows batch files
<teepee>
no, but in general ctest + python should work on windows too
Guest15 has joined #openscad
<teepee>
or lets say unlikely, who knows it might be possible to craft something similar with batch files and some extra tools
<guso78>
not sure, how well windows powershell performs, but why changing a running system ....
<teepee>
I did not have the time to look at the actual code yet, but the question is if we can find a way working together
<teepee>
I've seen so many dead projects on github by now which is really sad - thinking if all that effort could have been combined
<guso78>
1k forks does not mean 1k dead projects haha
<teepee>
no, of course not. and it's not even specifically the forks
<guso78>
actually i would love too see openscad running with an python interpreter alternatively
<teepee>
there's a number of projects in various languages that have a readmy "the better openscad" and then like 48 commits over a couple of month doing some basic CSG stuff
<teepee>
I've never missed that for the projects I do, but I can see a number of interesting cases where that would be really neat
<teepee>
like getting data from a database or so
<kintel>
They also did a proper fix in -4, but it's not being picked up yet
<guso78>
as in evolution: only the best species survive ... and you know which version is most active ...
J23879218 is now known as J23
<teepee>
I don't think software works that way. or we would not have JavaScript at the top of uses ;-)
<guso78>
Haha, Javascript has the same martketing model like M$. M$ has its genes shipped with every new computer ... not neccessarily perfect
<J23>
total misconception that the best survives it is only the best adapted at that time (and sometimes there is some luck involved too)
<J23>
and mainly it is the "if it just works it is enough" model
<teepee>
yeah, but also "can be used easily by everyone"
<guso78>
yes, and offering it in windows instead of linux only is still the main key factor ....
<J23>
was speaking of biological evolution.. software has that lobby and advertised thing. I mean how comes that jpegXL (.jxl) is not used
<teepee>
it's probably an important one, yes. but also other things like documentation, tutorials, ....
<teepee>
or 3mf :)
<guso78>
personally i dont see any benefit from 3mf. IMHO 3mf contains so many extra attributes in the file. If program A supports set A of the attributes and Program B supports Set B of the attributes,
<J23>
.. hm my printer sounds like a donkey having sex .. iiihh ahhhh iiiihhh ohhhh
<guso78>
then the interoperatibility between these programs is nearly random. my interchange format must be simple and predictive like STL ....
<guso78>
haha, my printer could also need some oik
<guso78>
oik
<teepee>
STL is too simple
<guso78>
oil
<teepee>
I suppose 3mf does have a bit of featuritis, but I'm not sure that's a huge issue
<teepee>
being able to actually store clear geometry, and with 3mf-v2 at double precision is much superiour to STL
<J23>
i think 3mf is just badly implemented .. when i check what cura or prusaslicer generate as 3mf - that is just terrible
<J23>
also 3d builder windows can make terrible scaling with 3mf files
<guso78>
sorry, need to leave, guys... wife is calling
<J23>
ihh ohh ihh ohh (sorry pretty sure it is not lubricant causing that)
guso78 has quit [Quit: Client closed]
<teepee>
I think mine make the sometimes horrible noises because the base is not 100% square - which I wanted to fix for years :)
<teepee>
nice that MSYS2 build ^ is green \o/
<J23>
sound is just caused by rapid travel movements