<SiFuh_>
ukky: did some rewiring and light fixture installation next door. Then repaired the outside lights for two homes. Washed the truck. Tomorrow, I think I have a bed frame to build. Not 100% sure. They want to borrow my drill but this is one tool you can't borrow. I have to be there. Because I use it almost daily.
<SiFuh_>
ukky: 3 days before the water becomes not so cold but still cool
<farkuhar>
zorz: it's not much of an accomplishment to upset ppetrov^, if he really deserves the label "human bot" for having such predictable reactions when long-held beliefs are challenged. If you're saying he appeared to have greater tolerance for lavaball's other antics, maybe that's because the norms lavaball habitually violated were not the ones ppetrov^ clung to most strongly.
<farkuhar>
So all it would take for lavaball to upset a "human bot" is to start violating norms that are more deeply held by the human bot, rather than the norms that the human bot cares less about. But maybe lavaball wasn't trying to maximize the number of people he offended, and the extra effort to trigger a reaction from ppetrov^ wasn't worth it.
<SiFuh_>
farkuhar: I thought ppetrov^ handled it well even though lavaball was 100% correct
<farkuhar>
SiFuh: I actually don't recall the details of that interaction.
<SiFuh_>
farkuhar: Not to worry. Lavaball slaughtered ppetrov^ because lavaball was using rationality and ppetrov^ was preaching regurgutated nonsense.
<SiFuh_>
Covid became a religion. A religion of fools
<SiFuh_>
No they 7 US states want Fauci in prison because of it.
<SiFuh_>
No they/Now*
<ukky>
SiFuh_: I have to do a lot of soldering right now. Need to replace nonreplaceable fuse bulbs with real fuse holders.
<farkuhar>
In #crux-devel jue hinted at why the libxml2 version bump was delayed for so long. But seeing as I already have problems with the llvm,clang,rust toolchain, there can't be much harm in updating the libxml2 dependency first, and then slowly rebuilding the toolchain starting with llvm.
<farkuhar>
The nice thing about having so much of the toolchain in a musl overlay is that there aren't as many uncommitted changes in the official repos that need `git stash`. All I have to backup is my pending version bump of contrib/firefox, and then the sync with upstream repos can proceed.
<SiFuh_>
ukky: Yeah I have soldering to do soon too. After this batch of beer I need to change the charging board.
<farkuhar>
As it turned out, my CRUX-MUSL laptop did have a copy of the llvm#19.1.7 package built back in January. But even after reinstalling that older package and rebuilding clang, I still hit the libclang errors with one of the rust crates in the firefox source tree.
<SiFuh_>
ukky: fortunately it is cheaper to buy the complete board than replace the chip.
<SiFuh_>
I bought 10 replacement boards 89 Malaysian cents each.
<SiFuh_>
No need to program. Just desolder and resolder the new board. So I bought a couple of rolls of desolder tape.
<farkuhar>
I'm curious why jue only bumped libxml2 to 2.14.0, when there was a 2.14.1 released more recently to patch a small regression. I guess I'll update only to 2.14.0 and see what needs rebuilding.
<SiFuh_>
farkuhar: Ask him.
<farkuhar>
SiFuh_: It's not important. Whatever gets rebuilt after the ABI breakage from the 2.13.7 -> 2.14.0 update, should still work after the version bump to 2.14.1, whenever jue gets around to it.
<SiFuh_>
So don't be curious then
<SiFuh_>
ukky: By the way that super powerbank I was working on gets stuck at 65 and 69% and drains very very fast.
<farkuhar>
Hmm, it's not such a big list according to revdep ... ffmpeg, graphicsmagick, imlib2, librsvg, libsoup, libxkbcommon, libxslt, nfs-utils, python3-lxml, shared-mime-info all seem to be broken after the libxml2 update. Although revdep didn't detect breakage in llvm, I planned to rebuild it anyway. But at least now there are other broken ports I can work on fixing first.
<SiFuh_>
ukky: I guess it is the wrong chip for the amount of milli amp hours
<SiFuh_>
ukky: Here is one story for you. There is a Willy's Jeep dissassembled here. I want the Chassis and Engine number to verify if I can register it with the Transport department. My wife's uncle works beside that guys shop. I asked him to get the number for me. He refused but asked that guy's son. The son refuse saying "No way!" Hehe. Everyone is afraid to talk to him about it.
<farkuhar>
AFAICT my two most recent builds of llvm#19.1.7 (January 28 and March 20) were configured almost identically to the official opt port. After accounting for trivial whitespace differences, the port in my musl overlay only omits the flag -DLLVM_ENABLE_LIBCXX=OFF, which appears in the cmake command for the llvm in opt.
<farkuhar>
The use of CXX instead of C++ reminds me of the different orientations in which someone might be crucified. Wasn't there some Christian saint who refused to be crucified on an upright cross, and chose to have his cross tilted with two legs on the ground like an X?
<farkuhar>
The choice of CXX instead of C++ in cmake flags obviously arises from a different motivation than what inspired the Christian martyr (who presumably didn't want to usurp the unique role that Jesus played). Anyway, it's a curious parallel between two worlds that don't usually intersect.
<farkuhar>
According to https://llvm.org/docs/CMake.html, the cmake option LLVM_ENABLE_LIBCXX defaults to OFF, so presumably it should be harmless to omit that line from the llvm Pkgfile. But maybe that default is subject to change between llvm releases, and the devs decided to play it safe by declaring the value explicitly.
<farkuhar>
Another Boolean flag I might want to experiment with is LLVM_STATIC_LINK_CXX_STDLIB, which also defaults to OFF and has some interaction with the LLVM_ENABLE_LIBCXX flag. All these complications with static linking are why I suggested that remiliascarlet's expertise might be valuable in figuring out the correct way to build the toolchain on CRUX-musl.
<SiFuh_>
ukky: The other day I was taking tires out of the truck and Aunt Sally ran away.
<SiFuh_>
Today I was rolling them and she came and started nudging them with her nose.
<SiFuh_>
I was very curious and stepped back. I think she was helping me
<SiFuh_>
Well trying to help
<SiFuh_>
ukky: I had to speak to drown out farkuhar's spam :-P
<farkuhar>
https://llvm.org/docs/CMake.html lists 66 Boolean flags. If I had access to a build farm with 2^66 identical computers running CRUX-MUSL, I could ask each of them to try building a different configuration of llvm, and see which one eventually succeeds in compiling firefox. But without access to that kind of hardware, I turn to the #crux-social channel seeking advice, only to be called a spammer.
<SiFuh_>
ukky: I had to speak to drown out farkuhar's spam :-P
<SiFuh_>
farkuhar: This is crux social not crux-musl
<SiFuh_>
social = where crux guys hang out and talk shit
<SiFuh_>
It isn't even an officially recognized crux channel
<SiFuh_>
The problem isn't the question, the issue lies with your 4,000 lined paragraph that erases everything anyone said before you.
<farkuhar>
Although a build farm with 2^66 separate computers would do the task in parallel, the approach that's available with my actual hardware is much slower: rebuild llvm with one of the 2^66 possible combinations of Boolean flags, see if it fixes the firefox build, and if not, proceed to the next possible configuration. Given that it takes at least 40 minutes to build llvm on my current hardware, that option is unacceptable.
<SiFuh_>
I doubt many people even bother to read what you type. It is like "Oh farkuhar. Skip"
<SiFuh_>
No one is here to read the bible.
<SiFuh_>
I must admit. 6 lines of text, I just scroll past it.
* SiFuh_
away fuck you all
<dlcusa>
I, for one, would like to see a crux-musl channel.
<dlcusa>
Meaning I would join AND read.
<farkuhar>
"voluntary flight in to and out of DCA has become an intelligence test." <- So if I want to accept SiFuh's invitation to a road trip in KL, I ought to arrange my trip to start from a different airport. That or plan a trip using a surface-based mode of transportation.
<ukky>
SiFuh_: yeah, that's probably the problem, the chip has to know what is the total power of the powerbank. And if charging board is cheaper than IC, of course just replace the board.