We should backup the contents of the kernel directory, and restore it if it fails... because we'd be stuck without glibc being updated, and other software that depends on the kernel headers??? I'm especially peeved by kernel 2.6 not having a make dep command.. and being told by another software package to do a make-dep Reproducible: Always Steps to Reproduce: 1. 2. 3.
Well, this is probably one of the reasons why reason genkernel does a "make mrproper"... > because we'd be stuck without glibc being updated, and other software that depends on the kernel headers??? Headers are not sources. Headers are in sys-kernel/linux-headers which need no compilation.
not when the nptl useflag is enabled :O I don't remember what packages require the kernel source to be made before installilng.. but there are some! (glibc with nptl (native posix thread libraries) I've also been messing arround with genkernel a bit, and have it start and stop often... once the intial part is done.. So i get this worse than most people... but it can grow to a problem for more people in the future.
Well, patches are welcome but I don't see how aborting "genkernel" does anything special and wipes things out - it's no different to halting a normal "make dep" which would create issues as well.
well, you're right of course... just got an idea... can there be a requirement for "kernel sources make deped" for some software? and some warning when we emerge something, that our kernel sources arn't make deped *before* portage gets arround to actually building the package that requires it, and fails with a cryptic message? (which might actually include make dep... so it's not that cryptic.. just unfortunate that it breaks a long update)
Yes; modversions.h - 2.6 makes this on a build action such as bzImage and 2.4 requires a "make dep" for it.
well, you're right of course... this is kindof extraneous... I ran into this a few times and it took me a while to understand what was going on. Maybe genkernel can catch the abort signal and say _you will need to rerun genkernel --makedep to make sure that all software will work, or run make dep) It's not that hard to make a bash script catch control-c ... portage does it.. and I mentioned how to do it in.. the bug for detecting if genkernel was already running... (Tim took care of it with a ps though) just got an idea... can there be a requirement for "kernel sources make deped" for some software? and some warning when we emerge something, that our kernel sources arn't make deped *before* portage gets arround to actually building the package that requires it, and fails with a cryptic message? (which might actually include make dep... so it's not that cryptic.. just unfortunate that it breaks a long update)
Ok, I don't see how this is a genkernel bug; the issues you seem to be experiencing are due to a lack of modversions.h, most likely, which can be solved by running "make dep" for 2.4 or make {bzImage,modules,} for 2.6 or by just running "genkernel [args] kernel".
well, this bug basically came down to, genkernel should do enough tests before running make mrproper and make clean, to ensure that they will be created again... so there won't be mysterious bugs the other part of this bug is: that the software that breaks when there is no "make dep" step, should have a very clear error message on how to get it to work again.
