Seems as though there is a project that has taken the linux kernel sources and sanitized them for user-space. Perhaps this is a better choice for system headers than gentoo's current ebuild that attempts to sanitze the kernel sources itself. http://ep09.pld-linux.org/~mmazur/linux-libc-headers/ Reproducible: Always Steps to Reproduce:
Hmm, this looks like a neat project, but it'd take a lot of testing to see if these work. On the plus side, I am intrigued by some of the patches available, especially the iputils one.
this idea has been brought up and shot down in the past... i believe the quoted problem was that of maintainability issues. anyways, i like our sanitised headers... as of 2.6.6, things appear to just work at least on amd64 and ppc64. :)
I'm confused as to how these headers are not maintainable, somone else is in fact doing it for us, it seems far simpler than sanitizing our own headers.
the maintainer hasnt touched them in a while, and isnt maintaining them now. :) at least that's the explanation i got in gentoo-dev from a dev who contacted the maintainer.
*points to timestamp on latest file, 2.6.5 kernel headers* I'd say thats fairly new.
the linux-headers package doesn't attempt any kind of sanitation. It basically unpacks a kernel tree, runs oldconfig, for sparc system, make dep is also run, then copies include/linux and include/asm-<arch> to /usr/include (during the livefs merge phase of emerge). We still have various #ifdef __KERNEL__ defs in there and all, and so far, things haven't broken yet, which is what Lv was getting at. The actual process of sanitization, as I know it to be, involves processing all the headers to eliminate the __KERNEL__ ifdefs, and other things. There was some discussion on how to do this at one point, but it was decided that the current system worked best, and we'd follow along with the proposed userland headers concepts being proposed for 2.7.
the current status of linux-headers is pretty much perfect we have fixes, we give them to plasmaroo, life goes on the current headers have been proven from bootstrap and beyond (by myself at least) on x86/ppc/hppa/arm and i know others have done it with amd64/ppc64 perhaps we can look at this in the future but for now there's no reason to change kumba: i've filed a bug for myself with the iputils/iproute stuff Bug 52938
The other patches might be worth looking at too, and passed around to the maintainers of those packages.