IRC channel logs


back to list of logs

***Server sets mode: +nt
<damo22>youpi: i think you made a typo in hurd last commit, shouldnt one of them be machdev_argv not diskfs_argv ?
<gnu_srs>damo22: Seems like you have switched to porting rust. What about the rumpdisk and acpi stuff, is it on hold?
<damo22>i got stuck so found something else to chip away at
<damo22>but youpi committed something on sunday that might help
<damo22>im compiling everything again to try
<damo22>gnu_srs: rumpdisk is almost completed, but for some reason its not filling in /proc/6/stat
<damo22>youpi found something that needs to be called since glibc was refactored, but it doesnt seem to be working on my machine, maybe my glibc is too old?
<gnu_srs>youpis latest commit for hurd was for glibc-2.33, not available yet.
<damo22>checking for _hurd_libc_proc_init... no
<damo22>root@zamhurd:~# cat /proc/6/stat
<damo22>cat: /proc/6/stat: Input/output error
<damo22>root@zamhurd:~# ldd --version
<damo22>ldd (Debian GLIBC 2.31-11) 2.31
<damo22>gnu_srs: rust is a pain to port, i prefer to work on drivers
<damo22>someone has to implement the libstd crate for hurd which is basically porting all the system headers
<damo22>i agree with youpi that they should just interpret the existing headers instead of requiring a reimplementation
<gnu_srs>damo22: I was considering porting rust, but abandoned the idea. I think youpi made some efforts earlier on and gave up.
<gnu_srs>Porting ada to hurd was hairy; same problem as rust: Needs to be bootstrapped by itself. Solved by cross-compiling.
<gnu_srs>Next could be fpc (free pascal compiler) to cross-compile. That could maybe be much easier than rust.
<youpi>damo22: they said there was a compiler that does the work
<youpi>(see the comments in lwn)
<youpi>fpc is quite oldish and not maintained
<youpi>the new kid on the block is rust :/
<youpi>rust people said it's basically about generating from the headers with the help of the compiler
<gnu_srs>What about gmc?
<youpi>gmc ?
<gnu_srs>The modula 2 compiler
<gnu_srs>Maybe the name is different?
<gnu_srs>Ah its gm2
<youpi>I didn't know gcc had a modula 2 backend :)
<youpi>it's not used as a dependency for debian packages, it doesn't seem a priority :)
<damo22>rust can retarget a different os on the same ISA very easily but you still need the rust-libc
<damo22>since youpi already did the work in llvm
<youpi>I didn't really do the work in llvm, just refactored stuff :)
<youpi>that got broken at some point
<damo22>i think the actual compiling of rust will be easy
<damo22>we just need to port the headers
<gnu_srs>Can similar techniques as for gccgo be applied?
<youpi>yes, that
<damo22>what would i feed into ti
<damo22>/usr/include/hurd ?
<youpi>#include <stdio.h>
<youpi>#include <unistd.h>
<youpi>more precisely, the output of gcc -E on them
<damo22>where do i get the exhaustive list of the required headers
<youpi>basically it'd be the posix header list
<youpi>from susv4
<youpi>probably some of them are not needed
<youpi>the vxworks file actually says which files they got them from
<damo22>./target/debug/bindgen input.h -- --sysroot=./hurd-headers -I./hurd-headers/include -I./hurd-headers/include/i386-gnu -target i386-gnu >
<damo22>we should have a way to bootstrap <insert any programming language> from GCC
<damo22>otherwise we would need to provide the initial binary toolchain for every language
<damo22>for every platform...
<damo22>isnt that already a big effort to maintain for GCC?
<damo22>i mean, if i succeed to compile rustc for hurd, then what? we wouldnt upload a binary rustc to debian right?
<damo22>unless its reproducible
<gnu_srs>I have a vague memory that gcc is considering a rust frontend. Maybe I'm wrong. Anybody?
<youpi>damo22: once we have a rustc binary we can compile the rustc .deb package, and then rebuild it against itself, and then we can upload that
<youpi>depending on the languages, they are implemented in a subset of their own language, for which a C-based compiler exists
<youpi>e.g. ocaml does that
<youpi>but others don't, like ghc
<youpi>and then yes bootstrapping is needed
<gnu_srs>Found it:
<rekado>we have mrustc in Guix and use that to build the chain of rust compilers.
<rekado>we have mrustc in Guix and use that to build the chain of rust compilers.
***Emulatorman_ is now known as Emulatorman