IRC channel logs

<matrix_bridge><Jeremiah Orians> Nice to see progress with pnut

<siraben>trying to get qfitzach to bootstrap mes, but regenerating nyacc tables is the most expensive part

<siraben>so I might just have to target blynn-bootstrap

<stabbles>Has anyone ever played with the idea of reducing the total amount of source code in the bootstrap chain? The size of trusted binaries is pretty much minimal already, but you easily get GBs of sources. I've been thinking vaguely about something like creduce to trim the intermediate GCCs while maintaining a certain property like reproducibility of

<stabbles>the final GCC. Of course that's not entirely trivial, cause it's slow, there's a risk of introducing UB by deleting code, and you'd have to prevent deleting code redundant for one target but necessary for another.

<stikonas>stabbles: I think if you want to reduce amount of source code, you need to improve early compilers, not reduce GCC

<stikonas>but then again, depends on your bootstrap criteria

<stikonas>if you want to avoid pregenerated files

<stikonas>it becomes quite hard to reduce amount of source code

<stikonas>and in the end, modern systems are quite big...

<stabbles>I'm mostly thinking about C/C++/Fortran compiler bootstrapping while trusting the kernel, so just the sequence of ~3 GCCs. I think even if you bootstrap various autotools and other build tools, the sum of intermedaite GCCs will be the dominant set of sources.

<stabbles>but yeah, retiring the need for an intermediate GCC by writing a more feature rich C/C++ compiler with a shorter bootstrap chain is another way

IRC channel logs

2026-10-10.log