Summary: | media-libs/imlib2-1.4.5: toolchain-funcs.eclass: Kernel version could not be determined, please inherit kernel-2 or linux-info | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | gilhad <gilhad> |
Component: | Eclasses | Assignee: | Gentoo Toolchain Maintainers <toolchain> |
Status: | RESOLVED DUPLICATE | ||
Severity: | normal | CC: | axs, dwfreed, kfm, tschosie |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | x86 | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: | detect KV, workaround |
Description
gilhad
2012-09-27 23:57:00 UTC
I can confirm the same thing with at least qt-core and mpg123 (along with imlib2) and possibly more. So it's definitely not just an imlib2 issue. Still not sure what is causing it, but I wanted to chime in with what I have found so far. I'm continuing to look. This error message comes from this commit: http://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/eclass/toolchain-funcs.eclass?r1=1.117&r2=1.118 which is from bug #432390 As far as I can tell, this commit breaks every ebuild that uses tc-arch() or tc-arch-kernel() from toolchain-funcs without inheriting kernel-2 or linux-info. As a workaround until a fix is pushed, set KV to the output of uname -r when running portage, like so: KV=$(uname -r) emerge -NDuav world dwfreed: does not work for me :( But it is probabelly result of the bug you mentioned. I tried to do (in /usr/portage/gentoo/eclass/toolchain-funcs.eclass): + KV=`uname -r` local KV=${KV:-${KV_FULL}} and it worked for me (but REALLY slowed down the initial emerge phase with rotating dash -/|\) I have the same problem trying to emerge qt-core. The workaround from gilhad works (to put KV=`uname -r` into toolchain-funcs.eclass). But it slows down emerge. I hope this will be fixed soon, before a lot of users gets breakage :) media-libs/libmad-0.15.1b-r7: the same error. Created attachment 325174 [details, diff]
detect KV, workaround
I would suggest that Alphat-PC attach the amended patch to bug 432390 and that this bug be marked as a duplicate. As imlib2 is far from being the only package affected, there seems little to be gained in spreading the discussion across multiple bugs. It also breaks catalyst stage building Setting KV to uname -r isn't a good idea imo ; the whole idea behind this was to get the kernel version properly according to the eclasses that can do this (kernel-2 or otherwise). I added the ||die because the output of tc-arch-kernel is wrong/undefined when KV is unset, and this has been going on for a while. Things "just work" on x86/amd64 because the kernel config itself maps the (invalid/undefined) value returned by tc-arch-kernel to the correct one, but it doesn't for arches like ppc/ppc64 (ie mapping them to 'powerpc'). However, if it is pereferrable this could be changed to an eqawarn and not be fatal. I think making it non-fatal would be the better way to go than setting KV on 'uname -r' by default. Marking as duplicate, as requested. Also, just committed a fix to make the warning non-fatal and only be displayed when arch type check is 'kern' since that's the only time KV seems to matter. *** This bug has been marked as a duplicate of bug 432390 *** |