IRC channel logs

2024-12-15.log

back to list of logs

<stikonas>it's mostly gnulib that uses snapshots
<stikonas>and possibly a couple of other tarballs
<lanodan>Well github/codeberg/… autogenerated tarballs have the same kind of problem as any kind of snapshot though
<stikonas>true
<stikonas>though release tarballs have another problem, they are usually full of generated code :(
<lanodan>yeah :(
<Googulator>fossy: probably OK as a stopgap, but obviously not a full long-term fix
<stikonas>perhaps contacting Savannah people might make sense?
<stikonas>to ask why it was removed...
<stikonas>it might also be unintentional
<stikonas>e.g. happened during some upgrade
<fossy>yeah, that would be sensible
<fossy>if they did it on purpose i have no idea what a long-term fix would be :\
<fossy>hmm, actually we could do it through gitweb instead of cgit, they haven't disabled .tar.gz downloads there
<stikonas>fossy: are hashes different?
<stikonas>and what are the URLs, e.g. for gnulib?
<stikonas>I get project listing disabled here: https://git.savannah.gnu.org/gitweb/
<fossy> https://git.savannah.gnu.org/gitweb/?p=gnulib.git;a=commit;h=2e8033f55cbe31f49db6e78f2b9d15f2677bde90 e.g.
<fossy>ya, same checksums
<fossy>the urls are reallly ugly though, but whatever
<aggi>nice, after sanitizing musl-libc for linux-2.4 ABI and patching a few dozen builds almost everything seems to work: vis-editor, gdb, htop shows processes now
<aggi>with yaft terminal there's some decent unicode support
<aggi>that was a two-year hackjob to get it done
<aggi>lanodan: i am aware of the problems with coreboot relying upon gnu dependency graph
<aggi>ekaitz: however, even when risc-v support progressed for any type of POSIX that needs bootstrapping, if risc-v dependency graph was not sanitizied and continued to rely upon x86_32
<aggi>bootstrapping problems accumulate, because risc-v (or arm) rely upon x86 bootstrapping/crossdev plus all the problems that risc-v build-time dependency graph may or may not introduce atop
<aggi>hence it's moot pointing to x86 bootstrapping issues to justifiy risc-v if x86 could not be avoided anayway
<aggi>furthermore, even when UEFI firmware would hide 16bit real-mode with or without proprietary malice, then this doesn't mean the related bootstrapping issues of 16bit real-mode would be absent
<aggi>"you should only need an uefi bootloader"... that circuumvents the question how the the UEFI subsystem malice got introduced
<aggi>similar to 16bit x86 BIOS which remains an open issue indefinitely, UEFI just adds malice atop of it
<aggi>"you should only need an uefi bootloader"
<aggi>this doesn't solve any problems, it introduces even more
<stikonas>kpcyrd: hi, in May you had a question about ruby and pregenerated files
<stikonas>still interested?
<stikonas>I've spend a couple of hours on that and had some success
<stikonas>well, I've got toe the latest ruby, without too much difficulty...
<kpcyrd>stikonas: May is a while back, I don't remember the question. happy to hear the answer though. :)
<stikonas>question was here: https://logs.guix.gnu.org/bootstrappable/2024-05-10.log
<stikonas>and this is the way to avoid using those pregenerated files: https://stikonas.eu/files/bootstrap/ruby_bootstrap.txt
<stikonas>(I'm not claiming this is the shortest way but that worked for me)
<lanodan>Is there a reason for bison-2.3?
<stikonas>I needed some older bison
<stikonas>so I just grabbed something old from live-bootstrap
<lanodan>makes sense
<stikonas>by current bison 3.8 was not compatible with those
<stikonas>perhaps it created .c and .h files but they failed to compile
<stikonas>and I've just tried to skip 2.2 completely, but that doesn't work
<stikonas>ruby 2.7 does not build with ruby 1.8.5
<stikonas>well, 2.2 has this bison annoyance, 2.7 then has an annoying issue with openssl (it requires it but openssl 3 is not supported)
<stikonas>not really a blocker for bootstrap
<stikonas>but I just patched it out
<stikonas>3.0 and later ones had no issues at all...
<stikonas>so overall not too bad
<stikonas>and this was tested on ruby git repo, so hopefully on other pregenerated files
<stikonas>and on a system with no ruby
<kpcyrd>hm, I think there was more context why I was asking this question. I remember it was very directly tied to whatsrc. but cool that you've figured it out!
<kpcyrd>I think for a distro like Arch Linux it's eventually going to boil down to "can you bootstrap it from the bootstrappable-builds distro"
<stikonas>it might have been in the hindsight of xz
<stikonas>can we build directly from the repo...
<kpcyrd>yes, but I forgot why ruby specifically
<stikonas>probably other things too, just ruby is one of them...
<kpcyrd>since I don't really use it personally. might be due to something raised by some other Arch Linux person who was more involved with ruby (possibly anthraxx).
<kpcyrd>anyway, cool to know it can be bootstrapped!