IRC channel logs
2020-01-29.log
back to list of logs
<oriansj>fossy: For M2-Planet FILE is just an int <xentrac>the standard answer is that the fileno(3) function gives you the fd of a FILE* <oriansj>xentrac: thank you and it would be trivial to implement int fileno(FILE *stream); in M2-Planet (just return what was passed) <xentrac>as long as nobody needs ungetc, I think all the buffering stuff is just an optimization <oriansj>xentrac: well janneke did figure out how to do that in M2-Planet. <xentrac>oh! in that case you're home free :) <oriansj>xentrac: well yeah; we just need to finish mes-m2 and then we are done bootstrapping GCC from hex0 <fossy>when oriansj makes mes-m2 run mescc <oriansj>fossy: well the bigger half will be when I get mes-m2 able to bootstrap guile and run guix <fossy>personally im more excited about hex0 -> GCC <fossy>Copyright (C) 2002 Free Software Foundation, Inc. <fossy>janneke, you may be interested to know that i built that make from tcc-boot0 <fossy>;; FIXME: add sensible check target (without depending on make) <fossy>so you could hypothetically move make to before tcc-boot and depend on make <fossy>can someone point me to a link to gash-utils <fossy>hm, trying to construct tcc-boot... -D seems to be highly buggy <fossy>am plagued by errors like libtcc.c:757: error: identifier expected <janneke>fossy: beautiful -- looks like the hard bit is done! <fossy>still fiddling with tcc-boot tho <fossy>im tempted to just try doing the whole bootstrap without tcc-boot lol <janneke>fossy: i have done many circles before the whole path worked <janneke>fossy: in the scheme-only bootstrap, that's what we do <janneke>tcc-boot0 builds make, gzip, bzip2, sed0, diffutils, patch, bash <janneke>tcc-boot is really just "nice to have": to build an unpatched tcc <fossy>janneke: well mescc-tools-seed dosent use tarballs it uses git submodules (for better or for worse) so we can skip gzip bzip2 and patch <fossy>janneke: how about the binutils glibc GCC triplet <fossy>is that built using tcc-boot0? <janneke>fossy: no, i'm using tcc-boot there; but not for a technical reason afaic remember <janneke>you probably really want to build binutils, gcc-core, glibc, gcc <janneke>and say you bootstrapped this gnu triplet <janneke>fossy: about mescc-tools-seed using git; that's just a developer's convenience <janneke>in my wip-full-source-bootstrap setup i use mescc-tools and m2-planet tarballs <jelle>janneke: I will mail you back soon btw :( I'll be at FOSDEM btw =) <jelle>I have an Arch day tommorow and friday =) <jelle>36 participants that's great <janneke>yeah, would be good to have some cross-distro event though; guess that's sat+sun <jelle>I probably won't attend talks so that would work <dddddd>But I'm looking forward the live stream. <dddddd>oriansj, I guess that the next step on the AArch64 front would be adding support to the cc_* family, right? <oriansj>dddddd: nope; after AArch64 is integrated into M2-Planet the steps have to be developed (hex0, hex1, hex2, M0, then finally cc_AArch64); along with porting MesCC to AArch64 <oriansj>A good step is writing the pieces in NASM (which we then can convert to M0 and hex*) <oriansj>But honestly I'd first get AArch64 into an M2-Planet release and probably do the stage0 steps for AArch64 after I do the steps for ARMv7l (as that would be easier to do) <dddddd>oriansj, I see. I'm hoping to send you the patches for M2-Planet not later than this weekend (past weekend I tried the commit-per-test but I didn't like it and prepared another idea of patch split). So, let's celebrate FOSDEM with the last(?) review and merge. (: <dddddd>I'm not quite conviced of the need of NASM versions (which can't be used in a proper bootstrap AFAIU) but if there's kind of pressure to deliver them... so it be. <fossy>dddddd: no they can't be used in the bootstrap but they can be used to verify the bootstrap and help people to understand the hex*/M* code <dddddd>You can get there disassembling the thing, which feels better for verification that looking to another (related, but not itself) the thing <oriansj>dddddd: actually the NASM allows one to develop the 1:1 assembly required for the tools without having to figure out the M1 DEFINEs you need to use <oriansj>So not absolutely required but I find it handy to get an idea before I write the M1 DEFINEs and do a line by line conversion to M1 (and then to hex2 and below) <oriansj>also I was hoping for a M2-Planet release before janneke's talk so he can announce the release at the conference <oriansj>(no rush fossy, I don't think you'll be able to get kaem good enough to bootstrap make via TCC or GCC in time for the talk) <dddddd>Indeed an asm is nice to create the definitions (I used gas/objdump to get the encoding). <oriansj>Plus I am aiming to get everything into talk-notes before thursday night, so that janneke has plenty of time to ask for clarifications.