IRC channel logs

<stikonas>rkeene: (cc oriansj) unlinkat on aarch64 is messed up: https://github.com/oriansj/M2libc/commit/9dc92269508651d7c4fdfa5a8f1456191cc868fa

<stikonas>asm( statement doesnt' even close function bracket ...

<lrvick>Can hex0 not run without a shell?

<lrvick>Calling hex0 with an exec and it just hangs. Calling it from a shell and it works

<lrvick>Specifically:

<lrvick>Hangs: RUN ["/hex0","/stage0-posix-x86-a15adc72f099506fb0d40c8824d7aaa97735f7c3/hex0_x86.hex0","hex0"]

<lrvick>However this works: RUN /hex0 /stage0-posix-x86-a15adc72f099506fb0d40c8824d7aaa97735f7c3/hex0_x86.hex0 hex0

<lrvick>The former uses an exec, the latter uses /bin/sh as a shell

<rkeene>stikkonas, Oops -- I think it was a merge error when cherry-picking

<lrvick>Guessing I actually do need kaem minimal...

<lrvick>ah needed to name my script kaem.x86. things are cooking now

<nimaje>lrvick: would the first one basically be execl("/hex0", "/stage0-posix-x86-a15adc72f099506fb0d40c8824d7aaa97735f7c3/hex0_x86.hex0", "hex0", NULL) or execl("/hex0", "/hex0", "/stage0-posix-x86-a15adc72f099506fb0d40c8824d7aaa97735f7c3/hex0_x86.hex0", "hex0", NULL) ?

<matrix_bridge><cosinusoidally> is the irc bridge down atm?

<matrix_bridge><cosinusoidally> Ah my message did appear on https://logs.guix.gnu.org/bootstrappable/2025-01-14.log but recent irc messages don't seem to be appearing in matrix

<matrix_bridge><cosinusoidally> I'm having another attempt at creating a simpler tcc bootstrap path https://github.com/cosinusoidally/tcc_simple not sure if I'll have the patience to finish it, but if I do it will be a case of cc_x86 -> tcc-0.9.27 in a simplified dialect -> stock tcc-0.9.27

<matrix_bridge><cosinusoidally> https://github.com/cosinusoidally/tcc_simple/tree/master/tcc_boot_min is a very cut down version of tcc-0.9.27 that is implemented in a simplified C dialect. It only implements that simplified dialect. That is complete.

<matrix_bridge><cosinusoidally> https://github.com/cosinusoidally/tcc_simple/tree/master/tcc_boot_max will be a cut down simplified version of tcc-0.9.27 implemented in the simplified dialect but capable of building a stock tcc-0.9.27. I'm at the very early stages of the port with tcc_boot_max

<matrix_bridge><cosinusoidally> tbh I might get bored and instead try to port chibicc to a simplified dialect of C and then use that to build tcc.

<matrix_bridge><Andrius Štikonas> I'll reboot the bridge later this evening...

<matrix_bridge><Andrius Štikonas> Hopefully that will fix sync from irc to matrix

<aggi>i'll try something else next, because i do want nilfs2 and 64bit timestamp filesystem support, and that's not easy to backport into linux-2.4

<aggi>tinycc-compilation/linking of some linux-2.6 series kernel

<aggi>i refrained from this for a simple reason, that was linux-2.4 is much less lines of code in total than 2.6 got

<aggi>concerning this fiwix would be ideal, 50K LoC with it, linux-2.4 got 4million LoC, linux-2.6 got 8million

<aggi>not sure yet, digging into nilfs2 and kernel api revealed a significant increase in complexity at the filesystem/page-cache/block-layer with linux-2.6 series

<matrix_bridge><cosinusoidally> Andrius Štikonas: thanks

<matrix_bridge><cosinusoidally> How much of those 8Mloc of lines in Linux 2.6 are drivers? Presumably the core of the kernel is much smaller?

<aggi>of cause, nonetheless the trend is undeniable, and it's a huge gap between a sane kernel such as fiwix and any linux-2.x

<aggi>besides drivers the filesystem stuff is scary, in particular the transition for ext2/3/4 etc. etc.

<aggi>i am thinking of another option, to introduce a 64bit timestamp variant of ext2 filesystem, and labeling this ext2x or something, which isn't compatible with regular ext2

<matrix_bridge><Andrius Štikonas> Haven't checked but presumably

<matrix_bridge><Andrius Štikonas> also a lot is networking

<aggi>i'll finalizxe the linux-2.4/musl-libc/500ebuilds stuff, it's working; and there's enough cleanup tasks remaining before thinking about a year2038 filesystem

<aggi>it's just i did wanted to fully migrate my development host onto the tinycc/linux-2.4 system myself soon, for extensive testing; and i formatted all my disks with nilfs2

<aggi>i had written a tiny segment verification utility for nilfs, that greatly increased confidence into what's stored on flash memory

<aggi>if i move my stuff onto any other filesystem that fiwix/linux2 got, i'll loose that

<aggi>nilfs2 is a killer feature for linux kernel

<aggi>now i remember: linux-2.6 compilation required some generated files produced from gcc -S, which complicated the whole approach with tinycc (or any other alternative toolchain)

<aggi>if i want 64bit nilfs2 back on board, i have to cope with any complexity added by linux-2.6 series, or have to live with what i got for linux-2.4

<aggi>nilfs2 was great for forensics too, because correctly administered it can prevent some secretly hidden filesyste i/o (otherwise i had not noticed there was some with some linux-5.x kernel)

<aggi>moving towards linux-2.6 has some other bad taste and smell attached, that was the DirtyCoW one which i've not forgotten

<mid-kid>made the mistake of deleting my distfiles directory again...

<mid-kid>Can't get flex-2.5.11, flex-2.5.33 and gc-8.0.4

<mid-kid>on the v1.0 branch

<lrvick>@nimaje should be former

<mid-kid>Managed to obtain gc-8.0.4, now flex...

<mid-kid>Found flex-2.5.33 in the debian archives: https://archive.debian.org/debian-archive/debian/pool/main/f/flex/flex_2.5.33.orig.tar.gz

<mid-kid>gc-8.0.4 was on archive.org: https://web.archive.org/web/20240518115554/https://www.hboehm.info/gc/gc_source/gc-8.0.4.tar.gz

<mid-kid>Still looking for flex-2.5.11

<mid-kid>(both of the above checksums match btw)

<mid-kid>Found flex-2.5.11 on a random ftp site: https://ftp-tel.sjtu.edu.cn/sites/ftp.openpkg.org/sources/DST/flex/flex-2.5.11.tar.gz

<nimaje>lrvick: well, I should have asked what argv[0] is, instead of assuming (wrong) what execl does, because I throught it would be easier to write the question that way, argv[0] should normally be the program name and is probably the easiest to get wrong when using the exec family of functions

<matrix_bridge><Lance Vick> rkeene: It would seem that symlinks extract, but directory structure is not preserved: https://dpaste.org/xVxgN/raw

<matrix_bridge><Andrius Štikonas> cosinusoidally: bridge restarted

<stikonas>testing

<matrix_bridge><Andrius Štikonas> looks like it is working

<rkeene>lrvick, I'll take a look tonight