<zorz>
farkuhar: forget the american stuff chatgpt, use china deepseek/qwenlm.ai/
<zorz>
SiFuh: I consent to give my data to China :P
<zorz>
yeah, valuable data hahahahhaa
<farkuhar>
trim_filename in pkgutil.cc is broken, when quotes are used to prevent word-splitting of the pkgadd command. For example, `pkgadd -r "/mnt/My Guest Filesystem" $my_built_package`
<farkuhar>
You get this error. pkgadd: could not read directory "/mnt/My Guest Filesystem"/var/lib/pkg: No such file or directory.
<farkuhar>
Or maybe trim_filename can be left alone. Just take care to strip away the quotes from the argument passed to db_open (const string& path) before initializing the variable "root".
<farkuhar>
This might be an instance of what SiFuh calls "failing to consider all eventualities". Per wrote the code under the assumption that nobody would ever need to pass a quoted path as the argument to the -r|--root option.
<farkuhar>
Eh, never mind, it seems to be an issue with the bash wrapper script, not pkgutil.cc. Prefacing the pkgadd command with `eval` is enough to overcome the quoting problem.
<farkuhar>
Still, I wonder if db_open could be modified to handle correctly the original command constructed by the wrapper script (before `eval` was prepended).
<ukky>
farkuhar: do you use grub or syslinux on your musl system?
<farkuhar>
ukky: definitely not grub. Either syslinux, or directly editing the boot entries in the UEFI.
<ukky>
farkuhar: ok, I will add syslinux to ISO's rootfs.
<farkuhar>
ukky: thanks for sharing your expertise for the ISO building effort.
<ukky>
farkuhar: too soon, I need to fix 191 failing builds
<farkuhar>
The core team is suprisingly silent, on the issue of libbsd and libmd in core (darfo) and the proposal to abandon httpup (SiFuh). But it's only been a day, so I guess we just have to wait patiently for their response.
<ukky>
I missed the reason why libbsd and libmd were added as dependencies to install ISO (or maybe dependency for some core package). But both libraries are now in 3.8 ISO's rootfs.
<SiFuh>
ukky: syslinux is the choice rather than efibootmgr because UEFI entries can vanish when disks change
<farkuhar>
The first reply to shadow issue #779 reads "What speaks against adding libbsd to your base system?" ... and then eventually jue goes ahead to do just that.
<SiFuh>
farkuhar: Yeah I was also wondering about that at the pub today. Why #crux-devel has replied to my httpup issue
<SiFuh>
farkuhar: ukky: When I start again maybe on Monday, I will look into lbbsd and libmd. If I think it is not relevant, I will probably remove it
<ukky>
SiFuh: Yes, System BIOS deletes UEFI enties for missing disks.
<farkuhar>
When libbsd was first added to CRUX (not in core), they tucked it away in a non-standard path (/usr/opt, I believe) so that it wouldn't be auto-detected during configure. But ports that needed to link with libbsd could still pass the includedir and libdir explicitly. I preferred that solution.
<zorz>
its funny, farkuhar does the work that beerman is supposed to do :P
<farkuhar>
Stop zorzing.
<zorz>
hahahaha
<SiFuh>
ukky: You are doing the 3.8 build?
<farkuhar>
beerman does the work that none of us care to do. He's following through on one of his responses to Fun in FS#1410: "why people might leave CRUX? 1. not enough ports (that work!) - many things aren't easily available in contrib that people might need to feel home."
<farkuhar>
By providing all those ports that help users feel at home, beerman is trying to stem the flow of dissatisfied users migrating to other distros.
<ukky>
SiFuh: Yes. From scratch, adding patches one by one.
<SiFuh>
farkuhar: 2. It became a totalitarian dictorship that ignores how the system should be and only a handful of people make choices without acknowledging what the users really want.
<zorz>
crux its ok with ports.
<SiFuh>
ukky: Don't forget to remove that stupid port in core dumb_runtime_dir.
<ukky>
SiFuh: Already removed dumb_runtime_dir.
<SiFuh>
ukky: and if you introduce sudo, move it to opt please
<SiFuh>
I've done everything except for glibc-introspective, json-c and httpup
<zorz>
hahah dumb_runtime_dir for login ooo g
<ukky>
SiFuh: For sudo, I do not care, it is not important where it is.
<SiFuh>
ukky: It shouldn't be in core. It isn't essential. su works fine. If you want sudo or opendoas or doas then sure, add it. Shove them all in opt.
<farkuhar>
SiFuh: That wasn't beerman's second argument. What he actually wrote was "2. we don't have people that actually sign up to care about our core tools, and thus, we are lacking a few features that we need as maintainers (git support in pkgmk, ..)"
<SiFuh>
farkuhar: Actually maybe that should be a new install option. Do you want to use an su helper? and gives you the option to choose between the three.
<ukky>
SiFuh: I know, it should not be part of the core, but this is political issue, you have to discuss it with crux-devel.
<zorz>
go there #crux-devel.......
<zorz>
:P
<SiFuh>
farkuhar: git should be opt as well httpup. But I understand that git is quite a useful tool. But being so, doesn't mean it needs to be in core.
<SiFuh>
ukky: linux-pam should be in opt as well. With a lengthy readme
<farkuhar>
beerman concluded with the third bullet point that I quoted in #crux the other day. None of his points really undermines what Fun was trying to say in FS#1410.
<SiFuh>
ukky: Does your script bootstrap?
<SiFuh>
ukky: Also will you be using the latest python3 that 3.8 CRUX is going to use?
<ukky>
SiFuh: I'm not there for 'make bootstrap' yet. Failing stage0. But I have not added _any_ musl patches to 3.8 yet. Even binutils fails.
<SiFuh>
ukky: Normal, I don't know how jaeger does it.
<farkuhar>
SiFuh: why even include httpup in the opt repo? If it's as unmaintained and broken as you said in #crux-devel, we could just drop it entirely, but still host the tools/httpup repo on Gitea (for historical purposes).
<ukky>
SiFuh: I will try to use everything from 3.8 branches, except linux-pam and that dumb thing (and glibc ofc).
<SiFuh>
When I built the Beelink version of CRUX, bloody hell that was the worst. It took almost 2 weeks to get a functional ISO.
<SiFuh>
farkuhar: Yes, good point
<SiFuh>
ukky: I don't even understand how that 'dumb' thing got into core. It is for a GUI and core is not GUI
<ukky>
SiFuh: It's some XDG_* whatever and some /run/lib or whatever. Those who use fancy desktop environments need it.
<SiFuh>
Exactly!
<ukky>
If people like MS-Windows desktop experience in Linux, they need that dumb thing.
<SiFuh>
ukky: Also I removed libxcrypt. MUSL doesn't need it
<SiFuh>
farkuhar: Methanol poisoning came back >15 so that is good. Still waiting on the MDMA which I doubt was the reason
<farkuhar>
FS#1111 (Stop using a hard-coded base directory in /usr/bin/ports) is another illustration that "we don't have people that actually sign up to care about our core tools". As reported by ukky, if the parent directory of a ports collection does not exist at the time of running rsync, then the operation will fail.
<farkuhar>
This failure mode (nonexistent parent directory) requires some deliberate deviousness to be triggered, but it's not impossible, especially for users who maintain an isolated installation at some mountpoint underneath /.
<SiFuh>
Doesn't /usr/ports exist under the port filesystem?
<farkuhar>
Both httpup and /usr/bin/ports take a very old-school approach to their respective operations. While /usr/bin/ports is more readable and maintainable, it's still slower than a compiled program like prt-get (on their common functionality, listing and diffing).
<SiFuh>
Oh ignore what I said. I was speed reading.
<SiFuh>
farkuhar: zorz: remiliascarlet: So the current plan is fly to Osaka, I drive through Iga and the mountains to Lake Biwa, then drive to Kyoto, then Hiba, then back to Osaka. Then we fly to Russia, and take the train to Lake Baikal, then from there no idea.
<farkuhar>
Expanding prt-get to cover the third function of /usr/bin/ports (syncing with upstream repos) would allow us to drop another historical artifact, consigning it to the archives along with httpup.
<SiFuh>
farkuhar: efficiency is the way. That was the per way.
<ukky>
SiFuh: In your notes: Remove all xorg entries. So, you are not building xorg for musl?
<SiFuh>
ukky: That was 3.7
<SiFuh>
for 3.8 xorg is done 100%
<ukky>
ok
<SiFuh>
Actually, you should have zero issues except xorg-xdm if I remember correctly ukky
<SiFuh>
xorg-xdm you need to remove linux-pam
<SiFuh>
Oh and for you slim users, there is another issue. You can't use slimlock without PAM.
<ukky>
We'll see, binutils done, 190 ports to go.
<SiFuh>
ukky: Not much needs patching to be honest. These days MUSL support works quite well. I was thinking though about firefox-bin. I know it is binary, but will it have issues? I never tested it.
<ukky>
Not sure if we can run firefox-bin as-is on musl system.
<SiFuh>
If it is truely independant, as in static. Then it should work
<ukky>
Pretty sure it uses shared libs.
<SiFuh>
It shouldn't need any libraries because it is a complete package and shouldn't rely on anything.
<SiFuh>
ukky: That is what I am concerned with.
<zorz>
SiFuh: Munich, this is the beginning of the end of NATO..... good step
<zorz>
Vance put auto-goal
<farkuhar>
joe9 in #crux is illustrating nicely beerman's claim "many things aren't easily available in contrib that people might need to feel home." Maybe a freeswitch port is being maintained in someone's private repo, but that's not much help for a newbie in joe9's situation.
<zorz>
farkuhar: joe9 illustrates why he should install debian
<zorz>
hahaha
<zorz>
decide what you want..... beerman tries to make crux debian
<SiFuh>
farkuhar: chinarulezzz and I were pissed when beerman was given repo access because with in two days, 80% of the ports were deleted.
<SiFuh>
So beerman can shut the fuck up
<zorz>
hahahhaa
<zorz>
pipx install magnet2torrent yessss
<ukky>
SiFuh: did you use argp-standalone?
<SiFuh>
No
<SiFuh>
I remember it came up 4 days ago, but somehow I didn't need it.
<SiFuh>
ukky: Can you show me the error. Maybe it will jog my memory
<ukky>
Do you use older (0.191) elfutils? I'm trying to compile 0.192 and it complains
<ukky>
SiFuh: configure: error: failed to find argp_parse
<SiFuh>
Yeah I saw that, but what was before it\
<SiFuh>
(0.191) elfutils
<ukky>
argp-standalone installs libargp.a
<ukky>
okay, 0.191 probably does not need it.
<SiFuh>
No the error did come up for me as well. But I can't remember how I stopped it.
<ukky>
3.8 uses 0.192 elfutils
<SiFuh>
I'd need your full log.
<SiFuh>
I am pretty sure I patched something.
<SiFuh>
When was elfutils 0.192 merged into CRUX 3.8? Because I was running updated of every port.
<ukky>
core-3.8 already has 0.192 elfutils, I'm using whatever is in 3.8
<SiFuh>
I was too
<zorz>
hahahaha the new captain america is black
<zorz>
farkuhar: good for you!
<zorz>
soon i will see Rambo VIII with black actor.
<SiFuh>
zorz: Wrong it will be a female transvestite overweight mexican
<farkuhar>
SiFuh: this is the junk food that dislodged the crown on my front tooth. https://0x0.st/8N45.jpg
<SiFuh>
farkuhar: Yuck, you humans and your sick confectionaries
<farkuhar>
As for "chewing more" as the strategy for an adult to regrow their teeth, doesn't it depend on the composition of one's saliva? Too acidic or too basic, and the extra chewing might be counterproductive for tooth growth.
<SiFuh>
farkuhar: Actually for regrowing enamel, that is saliva
<SiFuh>
For entire teeth that is something much different
<SiFuh>
And I don't remember ever saying "chewing more"
<SiFuh>
Speed reading are we again?
<farkuhar>
Okay, then what was your strategy for an adult to regrow their teeth? I can't be bothered to grep through the logs.
<SiFuh>
farkuhar: I think you are mashing different parts of the conversation into a single sentence to be honest with ya
<SiFuh>
Don't bother, because I never mentioned how to
<SiFuh>
farkuhar: two things in the body grow that do not require blood. Enamel isn't one of them. Enamel is a coating so it doesn't technically grow. Human teeth and cartlidge require the molecular level to shift molecules around. This can be simulated with low level ultrasonics.
<SiFuh>
To fix your crown, it won't work. To grow an entirely new tooth, it can.
<farkuhar>
Hahaha, every maintainer thinks that 80% of the ports available for the distro are bloatware and can be eliminated. Problem is, no two maintainers agree on which ports constitute the 80%.
<SiFuh>
It was discovered that if you use an ultrasound machine on your gums every night before sleep, you will regrow a new full set of teeth.
<farkuhar>
Actually, SiFuh and chinarulezzz might have been agreement on what constitutes bloatware. But their conception of the essential ports differed from beerman's, hence the reaction "pissed when beerman was given repo access".
<SiFuh>
He dumped much on contrib and opt.
<SiFuh>
Nothing to do with core
<SiFuh>
I don't give a shit what is in opt or contrib. I only care about the heart of CRUX which was core.
<farkuhar>
I should check whether darfo's ML post has gotten a reply yet. Somebody besides me ought to remember the non-standard install location /usr/opt that prevented eager linking with libbsd.
<farkuhar>
Still no reply from the core team, and eerie silence in #crux-devel too. Is it the calm before the storm, or just that everybody decided to go jungle for the weekend?
<ukky>
farkuhar: it would be funny when Crux dev team would port patches from crux-musl to fix libbsd issues
<SiFuh>
ukky: It certainly would
<SiFuh>
I very much doubt the high and mighty seated in their ivory painted towers of matchsticks would succumb to defeat
<farkuhar>
commit 3188225a110c1a3ef6beebac3c2796e514f442e3 in the contrib repo was jaeger's attempt to satisfy complaints like darfo's. I wish they would go back to that layout, and force the packages that actually need libbsd to adjust CFLAGS, LDFLAGS, LD_LIBRARY_PATH, and PKG_CONFIG_PATH themselves.
<farkuhar>
Heh, in this channel I'm just preaching to the choir. It would be more effective to follow SiFuh's lead, and post something in #crux-devel.
<SiFuh>
farkuhar: Yeah, grow some brass dingle berries and give them a swing in devel ;-)
<farkuhar>
SiFuh: Done. Let's see where the discussion goes from here.
<SiFuh>
I saw
<zorz>
farkuhar: besically why the need of #crux-devel hence no development in crux? :PPPPPppp
<zorz>
hahahaha
<SiFuh>
farkuhar: The envelope arrived at the clinic and it says private and confidential. I needed to blow up their bathroom. When I came back, the wife hands me the contents of the envelope. I looked at it and said "Cool, within normal" She asks "You understand it?" I said "Yep. it's less than 15 so it is in the normal range" She says "I had to ask the nurse to explain it to me" I said "You do realise you
<SiFuh>
both broke the law. It isn't for you. It is for me."
<SiFuh>
They are not legally allowed to open or give my medical information to anyone, including my wife.
<SiFuh>
Fortunately, I don't give a fsck. But it was funny to see the look on her face
<farkuhar>
You can bet that neither Alpine nor Void will have a patch for httpup (to prevent the random build failures).
<SiFuh>
ukky: jfsutils
<SiFuh>
farkuhar: They don't because none of them are as old as CRUX using dinosaur software.
<ukky>
SiFuh: cool
<SiFuh>
ukky: Yes, very cool. I actually enjoyed reading the code and screaming loudly as to why they are still using sprintf insteaf of snprintf
<ukky>
I like dinosaur software (if it works as expected).
<SiFuh>
ukky: Yeah... fluxbox!
<ukky>
Yes, fluxbox, I run it on NetBSD too.
<SiFuh>
farkuhar: ukky: It is amazing that even after many GCC upgrades/updates and library changes. fluxbox is till rocking along.
<SiFuh>
ukky: I started using it in 2003. I hate every other WM since. I only and still only use fluxbox.
<SiFuh>
ukky: I haven't looked at the code for fluxbox. But it must be rock solid since it isn't even maintained anymore and still compiles flawlessly
<ukky>
After using blackbox, not that many WMs are on par. Fluxbox is good match. Openbox sucks.
<SiFuh>
Last Updated 2016-07-30
<farkuhar>
There should be a distro policy that for every port added to core, at least one core port has to be dropped. That way you never have a core repo expanding beyond the initial size.
<SiFuh>
Yeah, I hated OpenBox
<farkuhar>
Exceptions can be made, of course, when a single port gets split into two, like musl libc and kernel-headers.
<SiFuh>
farkuhar: The only policy should be. If it isn't needed (as in essential to a functioning core) it should be in opt.
<ukky>
Same thing with 'runit'. I believe it was last updated around 2012-2014, but it rocks. SysVinit sucks compared to runit.
<SiFuh>
I like runit
<SiFuh>
farkuhar: Hence sudo should be opt. You don't fscking need it unless you want it.
<SiFuh>
Unix and Linux survived for decades without sudo. It doesn't need it. It's an optional choice so put the damned thing in opt
<farkuhar>
Drop at least one port for every port added, that's an easy policy to verify that devs are adhering to. It forces them to revisit old assumptions, and question whether each core port is worth keeping. Imagine what would have happened to httpup if such scrutiny had been applied port-by-port.
<SiFuh>
farkuhar: I understand the concept, but it won't because core ports should be neccessary ports.
<ukky>
I just added two ports to musl-core. Both are needed to compile elfutils.
<SiFuh>
Better idea farkuhar, buy me a ticket to Germany, and I will convince him to stop fucking adding shit to core.
<SiFuh>
ukky: What two ports?
<ukky>
SiFuh: argp-standalone and musl-fts, from Alpine.
<farkuhar>
SiFuh: two core devs are in Germany, but I assume you're referring to beerman, not jue.
<ukky>
You don't need them if you compile older elfutils (0.191).
<SiFuh>
musl-fts is needed for other shit too dude
<SiFuh>
and musl-obstack
<SiFuh>
No I like jue.
<SiFuh>
Known him for years.
<SiFuh>
So assume right ;-)
<ukky>
SiFuh: musl-obstack was already in crux-3.7-musl
<SiFuh>
I have both obstack and fts in 3.7
<SiFuh>
I added musl, musl-fts, musl-obstack and linux-headers to core because they were essential. Not like Linux-Pam, sudo and dumb_runtime_dir
<ukky>
I don't have 'fts' in my musl-3.7 ISO
<SiFuh>
ukky: Sucks to be you ;-) hahaha. How? I remember it was needed for building another core port
<farkuhar>
Why is frinnst not joining #crux-devel to help fix the libbsd mess? In #crux he might as well be shouting into the void.
<SiFuh>
argp-standalone <-- you won't need that with my patch.
<SiFuh>
But if I remember correctly another port also needed it. So rather patching two ports it may be a better choice to introduce argp-standalone
<SiFuh>
farkuhar: technically he is retired
<SiFuh>
He and I were the fluxbox guys. When beerman dropped fluxbox, I took the port over. Stupid to be honest to drop it. It never needed to be maintained
<farkuhar>
Okay, frinnst in the #crux channel gets at least one follow-up (from darfo), but for greater effect he should have left his comment in #crux-devel.
<ukky>
fsck me, gcc fails after successful build just because some 'rm' command from Pkgfile is deleting files that do not exist in musl build.
<ukky>
At least gcc takes only 35 minutes to build, not hours.
<farkuhar>
Although with this cruxbridge thing, maybe beerman is still seeing the chatter in #crux. Hard to tell what might eventually cause them to change course.
<farkuhar>
SiFuh: it's not libdm, it's libmd (message digest).
<SiFuh>
Okay done. ;-) I have titanium balls by the way
<SiFuh>
ukky: Yes. I commented those out
<SiFuh>
ukky: It took me three days to get everything to compile and I hated how I had to start from scratch everytime it would fail.
<ukky>
SiFuh: instead of commenting out, I will add ' || /bin/true'
<SiFuh>
farkuhar: Will be interesting to see the differences between what ukky and I have done. I bet we are very similar.
<SiFuh>
ukky: Sure. But it is MUSL, so you don't need them at all.
<SiFuh>
ukky: If CRUX accepted a MUSL version, it wouldn't be to difficult to have the ports glibc/musl friendly in the small crap like that.
<SiFuh>
We could have every port detect if musl or glibc and have slightly different configurations in ports.
<farkuhar>
SiFuh: it sounds like your ISO Makefile is more conservative than ukky's, hewing as closely as possible to the official ISO build. ukky was not constrained by any impulse to follow the same recipe as the official build.
<SiFuh>
Who the fuck wrote dumb runtime dir?
<ukky>
I am VCS-ing all my changes, it will be easy to post changes (except for content of patches)
<SiFuh>
farkuhar: I made the crux-3.7 following the official way crux is built.
<SiFuh>
farkuhar: emmett1 didn't so the system would break easily. He was great at helping out but he took so many shortcuts.
<SiFuh>
I told him, it needs to be official. So once we had done all the ports. I built it to the official standards.
<ukky>
My ISO Makefile can build glibc, musl, and stage0 (full toolchain ISO). It also allows non-root build of the first pass.
<SiFuh>
Creates an XDG_RUNTIME_DIR directory on login per the freedesktop.org base directory spec. Flaunts the spec and never removes it, even after last logout. This keeps things simple and predictable.
<SiFuh>
So again. WHAT THE FUCK is that doing in core?
<farkuhar>
It's very common for a port in the core repo to have a reflector email address in the Maintainer line, rather than a specific individual. dumb_runtime_dir is not unique in this regard.
<farkuhar>
If you want to see who actually put it into the core repo, just check the git log to see who made the initial commit.
<SiFuh>
Well, I should find out how did it and call Theo so we can punch them in the face for every syllable to pronounce the port name. DIR will great. That's 3 syllables
<SiFuh>
It's kind of cool. 1, 2 and 3 ;-P
<SiFuh>
And if they don't understand. one for each. "not, in, core!"
<farkuhar>
Does Theo even get involved in distro fights outside the *BSD universe?
<SiFuh>
I think so
<SiFuh>
But I am sure he would love to join me when punching a linux dev guy
<SiFuh>
remiliascarlet: Have you watch Idiocracy yet?
<SiFuh>
remiliascarlet: I asked you to watch it a few months back. You really should.
<SiFuh>
farkuhar: Interesting fact. Beavis and Butthead, King of the Hill, Idiocracy were all created by a friend of Alex Jones.
<farkuhar>
Did SiFuh just promote beerman by calling him a linux dev? I thought SiFuh previously assigned him a lower rank on the totem pole.
<SiFuh>
farkuhar: Stop farkuharing me ;-)
<ukky>
farkuhar: FYI, I am not splitting libc into libc and linux-kernel-headers for musl-3.8 ISO.
<SiFuh>
ukky: You are not calling it linux-kernel-headers I hope
<ukky>
SiFuh: For now, it does not exist. But I always wanted to separate libc in OS headers.
<SiFuh>
That would be like tautology in a package name.
<SiFuh>
You know like those fools that do /boot/efi/EFI/BOOT .... That is so dumb they should runtime dir with that. :-P
<SiFuh>
farkuhar: By the way, Colombian friend could fucking understand shit from the SYSLINUX documentation you provided. He ended up calling me.
<ukky>
I do not see anything wrong with that name. It could be replaced with hurd-kernel-headers.
<SiFuh>
couldn't*
<SiFuh>
I was on the bus to KL wanting to sleep and had to use a freaking phone to teach him how to install it correctly.
<SiFuh>
ukky: Linux IS a kernel linux-headers is fine
<SiFuh>
Besides most distros choose linux-headers and I have seen sime as kernel(version)-headers
<ukky>
Also, I understand that /boot/efi/EFI/BOOT looks weird, but you cannot mount UEFI boot partition to /boot mount point
<SiFuh>
ukky: I use /boot/EFI/BOOT/
<ukky>
Thats _very_ wrong
<SiFuh>
Why?
<ukky>
Your /boot needs to be FAT12/16/32
<SiFuh>
EFI/BOOT is fine for UEFI that is where it looks. And in linux I shove it under /boot/
<SiFuh>
Adding an second folder boot/efi/ is just stupid
<ukky>
Where do you place kernel? at /boot ?
<SiFuh>
In /boot/EFI/BOOT/
<ukky>
That makes your system the least reliable system in the world
<SiFuh>
Why does linux need a boot/efi to mount and EFI/BOOT?
<SiFuh>
Why can't I mount EFI under /boot directly?
<SiFuh>
/boot/efi <-- Linux EFI/BOOT <-- UEFI access
<ukky>
You don't need to mount EFI Sytem Partition at all. Only BIOS needs it at boot.
<SiFuh>
/boot/ <-- Linux EFI/BOOT <-- UEFI access MORE LOGICAL
<SiFuh>
Okay and why would I put my kernel in linux when I can do it in the UEFI access partition?
<ukky>
FAT is not reliable file system. You may endup with corrupt kernel.
<SiFuh>
So basically you are saying. If I add an extra empty folder as a mount point. It solves my problem of a corrupt kernel?
<SiFuh>
ukky: You know OpenBSD does exactly the same way I just said right?
<ukky>
If you use UEFI boot, you don't have a choice. It's a flaw in UEFI design that it can only boot from FAT partition.
* SiFuh
is looking for Theo. Oh wait, he lives in Canada.
<SiFuh>
ukky: No, it can book from ext2 as well. But not all
<ukky>
Never seen System BIOS that can read ext2.
<SiFuh>
book/boot
darfo has joined #crux-social
<SiFuh>
Many can.
<farkuhar>
SiFuh: darfo is here.
<ukky>
If it uses U-Boot, then yes, U-Boot understands ext2
<SiFuh>
darfo: farkuhar just told me you are here
<ukky>
SiFuh: You are just lucky.
<darfo>
indeed
<SiFuh>
ukky: Here is a fun fact, it can also support ext3, ext 4 and BTRFS depending on the manufacturer
<SiFuh>
ukky: And as for fat they can also support 12/16 and 32
<SiFuh>
fact*
<ukky>
SiFuh: I still have to see such System BIOS.
<darfo>
bell rings: "Bring out your dead!"
<SiFuh>
I repaired PCs for years and I saw many in my life. Hence, writing it in my documentation.
<ukky>
Well, I work in this area, and understand a bit or two about boot process.
<SiFuh>
ukky: I wasn't actually talking about you. I was talking more about the others in the channel reading this.
<ukky>
Maybe zorz programs EEPROMs in background and you don't know about this?
<SiFuh>
ukky: Oh hell no. I already spoke about EEPROMs with him and he had no clue.
<ukky>
SiFuh: okay. You can talk to me about that (if you ever get bored)
<SiFuh>
farkuhar: Actually this is one of the reasons I installed plan 9front. It has some cool ways to program EEPROMs
<SiFuh>
ukky: I don't get bored much. I have beer :-). You should see me when I am sober. I am like dead to the world. Like the Brain from Pinky and the Brain. Wife saw it once. It scares her.
<SiFuh>
farkuhar: Balls of titanium. Watch me when I am sober.
<SiFuh>
ukky: My wife understands that if I have beer, I am the most friendliest guy on the planet. Brain slows down so I can communicate with people. But off of it. I am in another world of my own and can go days without eating, just thinking and doing shit. Maybe I should sober up when doing 3.8?
<SiFuh>
ukky: The last time she saw me sober was when I stopped for two days because on day three I was to do wiring of an automobile. She said I was like Walter from The Fringe.
<SiFuh>
ukky: Our next trip to Japan will be interesting. I will need to be sober because I will be doing a lot of driving. This actually worries her.
<ukky>
SiFuh: Please build 3.8 after a few beers.
<SiFuh>
Hahaha. I was just thinking about the argument I had with the doctor. He refused to see me today. He lost face because I was right
<ukky>
Okay, attempt No.3 to build gcc
<SiFuh>
ukky: What machine are you building from?
<SiFuh>
Actually I was going to tell farkuhar about this.
<SiFuh>
farkuhar: I noticed if I use a dinosaur machine from before Adam was created everthing compiles fine. If I use an AMD 9 I run into many issues.
<ukky>
SiFuh: 2 CPUs, 48T total, 256G RAM
<SiFuh>
ukky: fuck....
<SiFuh>
CRUX needs more than 1GB to boot these days of RAM
<SiFuh>
ukky: When building 3.7 I did think of an idea. What if we build the ports for MUSL and use the GLIBC version to boot. But provide a prebuild kernel as well?
<SiFuh>
Funny how ukky talking calls to the forefront ideas I had in the past ;-)
<ukky>
SiFuh: I don't know if it worth it. It's not difficult to build full musl ISO.
<SiFuh>
Says you now ;-). I beg to differ after 4 days. Jaeger built it. We just replace the packages with MUSL packages. :-P
<SiFuh>
Introduce a prebuilt kernel into the install. That wouldn't be difficult.
<ukky>
And it is easy to shave some RAM requirements for bootable ISO is we remove linux-firmware and provide it as separate package, installable at boot.
<SiFuh>
Actually I was surprised linux-firmware existed. Usually that is downloaded rather than being supplied directly from most Distros
<SiFuh>
Thought it was a licensing thing
<ukky>
That's because you cannot distribute it freely (probably), but users can download it.
<SiFuh>
ukky: My mother is quite fermented so she now needs to do a medical each year to be able to drive. Freaking hell man.. I was listening to her story today. She had to write numbers into a circle to make a 12 hour clock face and draw hands that point to 11:10. Must be done within 30 seconds. WTF?
<ukky>
On the subject of injecting musl-packages into existing ISO: It might not be the same packages as built by system/iso/Makefile
<SiFuh>
That Makefile builds the packages on your current system. Installs them into a fake root, and rebuilds them, then repeats. You can do that easily without the need of a fakeroot or Makefile
<SiFuh>
I like the concept of a fake root, but I'd be more precise being as OCD as I am. I'd prefer to build it on your system. Create an ISO. Reboot and install, then rebuild it again. Create a new ISO, reboot and install and then rebuild it again and create to offical production ISO.
<SiFuh>
As tedious as that sounds, you or I as the builder would have a bit more control over what is happening.
<SiFuh>
But who am I to complain. jaeger has been building offical ISOs his way for decades.
<ukky>
I prefer fakeroot/container/chroot.
<SiFuh>
ukky: Yesterday I mentioned a problem with the script. Did you read it?
<ukky>
Yes, I've read, but which problem exactly?
<SiFuh>
If you control + c out of any of the stages it can't unmount the fakeroot and ends up deleting you entire work.
<ukky>
I have complained to jaeger about that in a few months I joined Crux
<ukky>
My current Makefile fixed that problem
<SiFuh>
That is why I back up the complete ISO each part. make kernel... backup. Make stage9 <-- modified for me. back up.... make stage 1. back up ... make stage2.. backup
<SiFuh>
Yeah I complained to jaeger about in 2019
<SiFuh>
stage9/stage1
<SiFuh>
err stage0 sorry
<ukky>
And I build stage0 as fakeroot.
<SiFuh>
It isn't in the text file from my repo provided. But I separated all the stages.
<SiFuh>
I made it so I can do make bootstrap0 and work soley on it.
<SiFuh>
Then make bootstrap1 and make bootstrap2
<SiFuh>
I didn't want the entire process to be automated through all stages
<SiFuh>
It's a time killer. bootstrap 1 fails and you need to go back to the start
<ukky>
Yeah, it needs manual control. When stage0 fails, I specify which single port to rebuild so it does not restart from square one.
<SiFuh>
I didn't see the point of bootstrap 0 and 1 being successful and then 2 fails. So you need to go back to 0 again.
<SiFuh>
It is a simple patch. Sure I have it laying around somewhere here. You can still do make bootstrap if you want. But having the ability to run specific stages manually was much better
<ukky>
chroot may fail because it might miss some file, but stage0 takes missing file from host.
<SiFuh>
ukky: Anyway. That Makefile script needs to be rewritten
<ukky>
jaeger might be okay with Makefile as-is
<SiFuh>
He runs it weekly
<ukky>
'Works for me' what he would say if we ask for changes
<SiFuh>
jaeger is awesome except linux-pam. But he would at-the-least give you his time to listen. And, if he likes the patch, he will probably introduce it.
<SiFuh>
Jaeger is the second longest CRUX user. I am the longest. So both he and I know per
<ukky>
gcc is done
<SiFuh>
gcc is actually the easiest part.
<SiFuh>
It's the smaller annoying packages. Also are you using patch?
<SiFuh>
Make sure you make a dependency for the very first port that needs it. I think rdate.
<zorz>
tcc
<zorz>
:P
<SiFuh>
Because when you do stage w and 2 it will error out saying patch not found.
<zorz>
trans cc
<SiFuh>
So I made a line Depends on: patch in rdate to cover it.
<ukky>
SiFuh: I will check that
<SiFuh>
There is probably a better way. But it works.
<SiFuh>
ukky: Yes, and seriously. Make sure patch is one of the first ports installed or when needed.
<SiFuh>
Without it, many ports will fail
<ukky>
SiFuh: patch is in core, thus it must be pre-built in stage0 (all ports in core)
<SiFuh>
Yes but when the script for Makefile does the pre-ordering it doesn't put patch first before the ports that need patching
<ukky>
SiFuh: It is already built: 101020 Feb 14 23:13 patch#2.7.6.17-9c98-1.pkg.tar.xz
<SiFuh>
So the very first port which I think is rdate you make it a dependency so it is installed before the port gets patched
<SiFuh>
If not the port will error out because patch doesn't exist yet.
<SiFuh>
ukky: It gets built in stage0 but in 1 and 2 it NEEDS to be installed before any of you ports that need patching are built.
<SiFuh>
ukky: In the old days, I use to control+z inject it into the fakeroot. But I found adding it the very first port as a dependency made the Mafefile script install it first during the sorting of dependencies.
<SiFuh>
zorz: WTF are you on about?
<SiFuh>
I am 99.999% sure the first port for patching is rdate
<SiFuh>
Edit the Pkgfile and make it a dep and it will install patch first before building rdate.
<SiFuh>
ukky: I was also thinking what if patch was put before rdate in the Makefile. Could that work? Never tried. Didn't want to waste my time
<ukky>
SiFuh: I will wait until build where patch is required breaks, then I will see how to avoid this in the future.
<SiFuh>
zorz: Not sure if you are a long time Linux user. But fuck gcc and glibc took forever to download on dialup and longer to compile.
<SiFuh>
ukky: Okay, just warning you to save your time. But pretty sure the first error is rdate. You did an rdate patch right?
<ukky>
I am newbie compared to everybody in this room, using Linux since 2012
<SiFuh>
Oh shit. Me since MIT
<SiFuh>
2001
<SiFuh>
Maybe 2000
<SiFuh>
I remember the first one though. Caldera
<ukky>
I was long time DOS/WIndows user. MS Visual Studio 2012 forced me to move away from Windows.
<SiFuh>
Boastingly, I was not. It was PC DOS, then BSD 1 and 2 then I use to poach my fathers win95 PC then straight to Caldera
<ukky>
SiFuh: I have rdate, but I didn't patch it for musl specifically: 18860 Feb 15 04:45 rdate#0.13-1.pkg.tar.xz
<SiFuh>
ukky: No it is patched for CRUX not for MUSL
<SiFuh>
I think it was NTP leap year or some shit.
<SiFuh>
ntpleaps.patch
<ukky>
There is ntpleaps.patch for Crux in core-3.8
<SiFuh>
I have
<SiFuh>
And my port is from 3.8 version 0.13
<ukky>
My first stage0 build had _no_ musl patches at all, I just wanted to see what fails.
<SiFuh>
The patch is written as #FS1462
<SiFuh>
ukky: Mine too
<SiFuh>
It is how I figured out what needs patching.
<SiFuh>
Nice to see you think along the same lines as me ukky
<ukky>
SiFuh: I had 173 failed packages
<SiFuh>
Many will be dependencies on a port that needs patching
<SiFuh>
When you figure it out you will probably need to patch less than 10
<SiFuh>
Also if you get a basename error. That is because MUSL doesn't seem to announce the base name.
<ukky>
Correction, 152 failed packages
<SiFuh>
The base name error is a super simple trick. Just need to introduce a MUSL header
<SiFuh>
#include <lingen.h> into to whatever .h or .c file requests the basename.
<SiFuh>
err
<SiFuh>
#include <libgen.h> into to whatever .h or .c file requests the basename.
<ukky>
thanks
<SiFuh>
Actually you should see two entries for that in my jfsutils patch
<SiFuh>
ukky: I was actually surprised that xorg didn't need any MUSL patches
<SiFuh>
I think libdevinput or whatever its name was needed a patch that was under xorg I think.