09:39 <abu[m]> Too bad I did not implement that from the beginning

09:39 <abu[m]> Opaque pointers are really a much better (= simpler) concept

09:40 <abu[m]> I can compile in LLVM 14 if I omit "-O3" from 'opt'

09:40 <abu[m]> And: I can make changes incrementally, so I can test step by step

09:41 <abu[m]> But: Bad news is that it is not only Makefile and src/lib/llvm.l, but many files in src/*.l too because they contain pointer types

09:42 <abu[m]> So it indeed needs two different src/ directories if we want to support both old and new versions

09:50 <tankf33der> ok

09:51 <tankf33der> i will do my best.

09:51 <abu[m]> Thanks!!

09:52 <abu[m]> I think opaque pointer work only at >= 14

09:52 <abu[m]> What to do with old LLVMs?

09:52 <abu[m]> Two src/ directories?

09:53 <abu[m]> Or wait a year or two, and then deprecate <14

10:04 <tankf33der> user base is so small

10:05 <abu[m]> right

10:07 <tankf33der> and no one could get llvm15 for their linux distro

10:07 <tankf33der> only rolling distro will be llvm15 soon

10:08 <abu[m]> llvm14 is also good (I use here on Termux and all my Debian installations)

10:08 <tankf33der> ok

10:08 <abu[m]> but no idea about e.g. llvm13

10:09 <tankf33der> now i think you should choose future and have only one opaque version

10:09 <tankf33der> you could split for older and maintain only opaque

10:09 <tankf33der> ====

10:09 <tankf33der> i could test on llvm13

10:09 <abu[m]> yeah

10:10 <abu[m]> You can test with this http://pb1n.de/?4cc63c

10:10 <abu[m]> on 14 it builds the non-opaque sources too

10:10 <tankf33der> asap, getting llvm13 ecosystem

10:10 <abu[m]> cool

10:14 <abu[m]> Does pb1n preserve TABs? (i.e. in Makefile)

10:15 <abu[m]> Anyway you know what it is. I made two globals OPT and OPQ to switch between old and new

10:15 <abu[m]> And -O3 needs to be off, otherwise 'opt' crashes with a segfault in llvm14

10:36 <tankf33der> i used opaque flag in clang too

10:36 <tankf33der> how it compiled without it?!

10:37 <abu[m]> No sources are changed yet

10:39 <abu[m]> And clang is irrelevant for us. It is used only to compile .c files or assemble .s files. The pointer stuff is in the stage .l -> .ll -> .bc

10:39 <tankf33der> clang used to compile lib.c

10:40 <tankf33der> all three must be on the same level of opaquness

10:40 <tankf33der> opt, llvm-link and clang

10:41 <abu[m]> ok, but on 14 nobody complained

10:43 <tankf33der> ok

12:17 <fuxoft> Sorry for the noob question but in the PicoLisp reference documentation, why do some of the definitions contain dot (e.g. "(de sym . any) -> sym") and others do not (e.g. "(diff 'lst1 'lst2) -> lst")? In both cases, I use these commands without entering the dots. Does this have something to do with dotted pairs?

12:18 <abu[m]> Yes, correct

12:19 <abu[m]> (de f (N) (foo N)) is the same as (de f . ((N) (foo N)))

12:19 <abu[m]> so 'any' is a function here

12:19 <abu[m]> Try to draw these expressions in box notation

12:20 <fuxoft> OK I'll try, Thanks for pointing me in the right direction

13:05 <abu[m]> If you can use Vi[p], you can try:... (full message at <https://libera.ems.host/_matrix/media/r0/download/libera.chat/63c7ebea50254c54070690d012ecfbd69ad715c4>)

13:06 <abu[m]> hmm, stupid IRC :/

13:07 <abu[m]> or, rather, stupid Matrix/IRC bridge cannot handle multi-line messages it seems

17:15 <fuxoft> Unfortunately, I've never used it. But I think I understand it better now. I need much more practice...

17:16 <abu[m]> Good ☺