01:03 <JordanBrown> InPhase: yes, I saw that reference. But if you follow it up, you find that that paper is on transfinite sets, which doesn't seem like the basis for the concept of normal sets.

01:04 <JordanBrown> And yes, MCAD is in poor shape. My discussion above was more about a hypothetical bundled library that one might actually want to use.

01:14 <kintel> pca006132 ugh, so I have to build tracy manually. They have a CMake config, a meson config, and the docs say just use the Makefile

01:54 <JordanBrown1> test

01:54 <JordanBrown> test

01:55 <JordanBrown1> (Experimenting with switching to Thunderbird chat.)

02:04 <JordanBrown1> test

02:04 <JordanBrown> test

02:05 <JordanBrown> Actually, I think I'll try Pidgin again. I don't remember why I stopped using it.

03:43 <kintel> pca006132 Got tracy working - turned out that the macOS version was a one-liner away. For the first time in a long time, the macOS version of a niche OSS tool was easier to install than Linux ;) ..and I had to run openscad as sudo to get samples.

06:15 <pca006132> kintel what one-liner did you add? And it is weird that it requires sudo for sampling, probably some OSX limitation

07:38 <JordanBrown1> Trivia: what is the output of this program?

07:39 <JordanBrown1> include = 0;

07:39 <JordanBrown1> if (include < 0 || include > 3) echo("out of range");

12:39 <InPhase> JordanBrown: I anticipated the keyword collision, but didn't anticipate the resulting behavior after handling the keyword collision. :)

12:41 <InPhase> I guess that's what you get from a text substitution command in a language with no semantic errors...

12:46 <kintel> pca006132 the sudo requirement was on Linux. On macOS I just "brew install tracy". I don't have a Linux build yet..

12:47 <kintel> -> so I'm executing openscad on Linux, but receiving traces on macOS, which was kind of what I hoped for anyway :)

12:48 <kintel> (only doing automatic sampling atm., perhaps that's what sudo was needed for

16:39 <pca006132> weird, it should not require sudo

16:39 <pca006132> but anyway glad that it works

16:40 <pca006132> you can use the ZoneScoped macro for instrumentation, it should give more accurate results comparing to sampling

16:40 <pca006132> e.g. https://github.com/elalish/manifold/pull/640

16:51 <pca006132> I tried adding opengl profiling to openscad, but without success, I think maybe because things like context creation and swap buffer are controlled by qt

16:51 <pca006132> don't have much experience with opengl so can't really debug that

17:16 <kintel> pca006132 do you know if I can get tracy to record stack frames too?

17:32 <JordanBrown> InPhase the thing that surprised me was that it wasn't required to be at the beginning of a line.

18:41 <JordanBrown2> kintel, teepee: lexer.l is inconsistent on whether it uses spaces or tabs, and when it uses tabs they seem to be set up for 4-space tabs. Got any opinion on standardizing it, to spaces, 4-space tabs, or 8-space tabs?

18:51 <InPhase> JordanBrown: That behavior persists back to the 2015 release (which only partially runs for me, but enough to test that).

18:55 <JordanBrown2> InPhase I'm not surprised; it seems more like the kind of behavior you would get when you are first trying to get things to work, rather than a deliberate feature.

19:27 <gbruno> [github] jordanbrown0 opened pull request #4891 (Ignore \r characters. Fixes #4882.) https://github.com/openscad/openscad/pull/4891

20:44 <gbruno> [github] jordanbrown0 opened pull request #4892 (Standardize lexer.l on 4-space indents using spaces. Align all actions.) https://github.com/openscad/openscad/pull/4892

20:45 <JordanBrown2> PR#4892 is a formatting-only change; it yields exactly the same scanner. kintel? teepee?

22:25 <teepee> JordanBrown: fine by me

22:26 <teepee> JordanBrown: isn't old macos also \r only?

22:29 <teepee> hmm, ok macos 9, I suppose that predates openscad :)

22:39 <teepee> https://stackoverflow.com/questions/50867928/how-to-handle-any-unix-dos-mac-text-file-in-flex-scanner ?

23:05 <JordanBrown> We could make it handle \r-only files, but as you say there are no OpenSCAD-capable systems that use that format, and really no recent systems at all.

23:06 <JordanBrown> I thought about it, and my conclusion was that ignoring \r was the simplest solution.

23:06 <JordanBrown> But if people want it to handle \r-only files I can make it do that.

23:06 <teepee> if the solution is as simple as in the linked post, why not just cover all the cases?

23:07 <JordanBrown> It's not that one pattern doesn't work. But I think there are five line-break-matching patterns in our lexer.

23:08 <JordanBrown> Though, you know, \r\n?|\n sure is pretty simple.

23:09 <teepee> and assigning that to NL or something would not make things difficult to understand

23:09 <JordanBrown> The count-lines macro would have to get smarter. (And sadly for only a very very limited case: include <> and use <> where there is a line break between the keyword and <.)

23:13 <JordanBrown> I was thinking that you would merge #4892 (the space/tab one) and then I would PR #4891 (the lexer fix) on top of it. That way somebody reviewing the changes can easily see the substantive changes in 4891 without them being obscured by the layout changes in 4892.

23:14 <teepee> sure, I can merge the whitespace fix if that helps

23:15 <JordanBrown> Seems like it leaves a cleaner history for those who follow.

23:16 <JordanBrown> Thanks.

23:16 <gbruno> [github] t-paul closed pull request #4892 (Standardize lexer.l on 4-space indents using spaces. Align all actions.) https://github.com/openscad/openscad/pull/4892

23:16 <gbruno> [github] t-paul pushed 1 modifications (Merge pull request #4892 from jordanbrown0/lexer-source-tabs Standardize lexer.l on 4-space indents using spaces. Align all actions.) https://github.com/openscad/openscad/commit/3d094bf87bbdc533b553169fcd7c0207774c3e22

23:16 <JordanBrown> I'll look at supporting the \r-only case.

23:16 <JordanBrown> But given its very very low value, if it's more than a tiny bit more complicated...