Bootable if you put in a GCC-compiled bootstrap object file, and exclude lots of drivers, and don't test things too thoroughly.
At least, according to the post here, and that only took two seconds to find.
It's like saying that the Linux build process is now usable without Perl... so long as you have Perl installed and compile a couple of kernel parts with it first, or do that elsewhere.
I have to agree: when "make" builds a working kernel from (even patched) source, then you have a reason to shout. Even self-hosting is a bit dubious if you haven't actually tested that the kernel you're using is stable and/or working even vaguely right except on a single example computer and a lot of hope.
When it's possible for *anyone* to compile a kernel without having to do more than apply a patch or two, (i.e. no more complicated than your average unsupported compiler) then they can rejoice. Until then, it's an interesting intellectual exercise that is *nearing* success.
Posted Oct 26, 2010 11:01 UTC (Tue) by tzafrir (subscriber, #11501)
[Link]
TCC did make headlines building a patched kernel.
Clang builds a working 2.6.36 Kernel
Posted Oct 26, 2010 13:23 UTC (Tue) by PaXTeam (subscriber, #24616)
[Link]
> I have to agree: when "make" builds a working kernel from (even patched) source, then you have a reason to shout.
then see http://lwn.net/Articles/409614/ ;). i posted a detailed email with the minimal patch last night to cfe-dev but apparently it's still held up in moderation, so i might bite the bullet and subscribe ;P.
Clang builds a working 2.6.36 Kernel
Posted Oct 26, 2010 13:59 UTC (Tue) by PaXTeam (subscriber, #24616)
[Link]
Posted Nov 6, 2010 3:08 UTC (Sat) by efexis (guest, #26355)
[Link]
"It's like saying that the Linux build process is now usable without Perl"
No it's not. If it says "without gcc" then you would be correct, but it never says "without gcc", it only says "with clang". An inclusive is not the same as an exclusive.