IRC channel logs
2023-10-07.log
back to list of logs
<[exa]>whew cool! just curious-- how long is the bootstrap chain? <h01ger>4km, or 327681337 stacked 3.5" floppy disks ;) <h01ger>i ment: in time, or in sw projects build steps? :) <janneke>[exa]: the bootstrap chain will only shorter when upstream projects hear about supply chain security and start taking their responsibility wrt being bootstrappable <[exa]>I meant-- iirc the latest gcc buildable with mes&pals was like 3.x, so I wondered how many gcc's are in the middle <janneke>in all honesty, you can *look* at most every compiler builder out there, including gcc and tcc <lilyp>Guix counts from boot0 to boot6 <janneke>also, if either gcc or tcc had taken their responsibility, this irc channel would most probably not have existed <lilyp>fair enough, but IIUC, both are at least reasonable enough to build using *a* C compiler <matrix_bridge><emilytrau> [exa]: the number differs depending on how you count it but i have about 30 packages in the chain between stage0 and gcc 13 <matrix_bridge><emilytrau> in terms of compilers my chain is: mes -> bootstrappable tcc -> tcc+mes -> tcc+musl -> gcc 4.6 no c++ -> gcc 4.6 -> gcc 7 -> gcc13 <matrix_bridge><emilytrau> i've seen other chains use versions like 2.95, 4.0, 4.7, 4.9, etc. i'm really excited that theoretically risc-v should be supported without needing to port the 4.6 patches to any other intermediary versions <janneke>lilyp: no, anything later than gcc-4.7 cannot be built with a C compiler; also there are many C compilers that cannot build gcc-4.7 <janneke>emilytrau: oooh, you beat us to it to remove gcc-2.95, well done!!! <janneke>ACTION has been working to do tcc+extended mes libc->gcc 4.6...but i guess that taking the musl route is a smart(er) hack <janneke>i agree that tcc + some-libc -> gcc-4.6 (with riscv, powerpc and whatnot backported) is the way to go <lilyp>should have been a C++ compiler :( <janneke>yeah, and we don't know of any decent "a C++ compiler"s other than gcc-4.6 that are bootstrappable <matrix_bridge><emilytrau> janneke: credit to t184256 for showing tcc+musl can bootstrap gcc 4.7 first ^_^ <janneke>+1, also for everyone helping eachother <janneke>emilytrau: chances are, that your bootstrap "just works" for arm too, we were having so many unexpected troubles (oabi/eabi) with gcc-2.95 and glibc-2.2.5 that making the jump from tcc to gcc-4.6 already seemed the smartest route <matrix_bridge><Andrius Štikonas> When working on live-bootstrap, I investigated various libc's <matrix_bridge><Andrius Štikonas> And musl is far easier to build than anything else <muurkha>thanks, that's really useful to know <muurkha>do you mean in terms of what it demands from the C compiler, or just its build system? <stikonas>build system is a fairly simple handwritten shell script + makefiles <muurkha>I suppose you don't have any idea how much RAM it needs, do you? <stikonas>but most likely far less than what mes uses <muurkha>I'm wondering if I can compile it with pcc on a microcontroller <muurkha>(the other, opposite meaning of bootstrapping: your computer can recompile its own systems software from source code, using a pre-existing compiled version of that software) <stikonas>well, musl wouldn't run on microcontroller <stikonas>rickmasters even had to patch it a bit to run on Fiwix