<oriansj>I might just create a rm for mescc-tools-extra just so that we can delete it out of principal <stikonas>though hopefully that file can just be deleted from mes <oriansj>yes it needs the unlink syscall but fortunately we will not need to support that syscall in builder-hex0 <oriansj>as the delete will happen with the next kernel <oriansj>well it is ultimately up to janneke what is included and not included in his mes tarballs <oriansj>I'll hack together a rm and add the syscall for M2libc <oriansj>assuming that I don't have to figure out the weird work around for riscv-32 again it should be done by tonight <stikonas>unlinkat seems to be supported on all arches on linux <oriansj>so yes, riscv-32 oncce again is missing syscalls <stikonas>well, I think most of those non-"at" file manipulation syscalls were made obsolete and replaced with "-at" versions, it's just that older arches have to keep them for compatibility <oriansj>and we have to use the older syscalls if we wish to be compatible with more POSIX kernels <oriansj>and libcs would have to provide a shim/work-around to support legacy programs as well <oriansj>so then I don't even have to do any M2libc work <stikonas>well, syscalls themselves are usually not visible to programs, libc's (even modern ones) still expose unlink function, it's just under the hood they are implemented using appropriate syscall <stikonas>we only need unlink() function in M2libc <oriansj>you even implemented it yourself for riscv-32 and riscv-64 <oriansj>and so rm.c is done and built by M2-Mesoplanet <oriansj>it keeps going after failures just like regular rm (even though it is probably a really bad idea) <oriansj>the build process is: M2-Mesoplanet -f rm.c -o bin/rm <oriansj>test with: touch foo; ./bin/rm foo; ./bin/rm foo; (first success and second failure with error message) <oriansj>and if you need it later adding rmdir appears to be rather simple <stikonas>but mescc-tools-extra is not becoming "mini-coreutils" <stikonas>we now have variants of mkdir, rm, cat, sha256sum, cp and chmod <oriansj>maybe just enough to bootstrap reall core-utils <stikonas>yes, real core-utils are not that hard to build <oriansj>and hex2 is like a better verison of od <stikonas>well mknod is primarily for /dev/null but having it is not essential <fossy>and i don't really think we need mknod even before linux kernel <oriansj>well we need GCC for Linux, so do we need /dev/null to make GCC? <fossy>no, we don't "need" /dev/null for anything, really <fossy>i don't think we ran into anything that cares whether /dev/null is a node or a file <oriansj>but would it make things more clean if we did? ***ChanServ sets mode: -o janneke