IRC channel logs
2019-09-12.log
back to list of logs
<OriansJ>Hagfish: nice find; it certainly looks like something that could be bootstrappable from M2-Planet with a reasonable amount of work. <OriansJ>Then one would be able to do the messy linux/gcc builds upon a posix; which really simplifies several minor techical problems. <OriansJ>Literally there are only 18 pieces left to complete the entire bootstrap (assuming janneke has finally removed the remaining issues from the MesCC to GCC path); that is if we include building trusted hardware (but not including ports to new architectures) <OriansJ>Here is my current (probably insanely stupid plan); I create a collection of scheme programs called slow-utils, that all can run on guile/mes-m2 that remove the need for mescc-tools and any other binary besides the scheme interpreter from our bootstrap chain. With the assumption of only mes-m2 [built by M2-Planet] aiming to build guile and gcc via MesCC. Everyone gets to mock how bad I am at programming in scheme and thus solves my need <OriansJ>for proper scheme test programs for the mes-m2 binary. Guix's bootstrap reduces to only the mes-m2 binary, which can simply leverage a kaem script to bootstrap guile and gcc <OriansJ>Thus eliminating the need for a fixed-point in the guix bootstrap binaries as M2-Planet always produces bit-predictable binaries and when I am done with mes-m2; all non-deterministic features will be stripped. *janneke has mes+mescc x86 segfaulting all of a sudden, terrible -> zZzz <OriansJ>janneke: I think me rewriting hex2+M1+blood-elf in scheme first probably is a good first step; then MesCC can just depend upon itself and import them as source and thus only depend upon it's interpreter (Thus able to ignore the environment entirely) <OriansJ>Then we can import gash, kaem, tar and other pieces to slow-utils until we have everything required for a full and complete bootstrap together.