1) Change in /etc/make.conf the CHOST to a valid GCC version 4 host name plus "-pc-gnu-linux" as specified under "-mtune" argument on the GCC web-site. 2) Do an emerge [-1] {linux-headers,svgalib,grub} You will get a bunch of messages that say ${tc-arch} has a value of "unkown" and failed. My case has CHOST="athlon-pc-gnu-linux". Reproducible: Always I have produced a patch which corrects most of the problem on x86_64/x86 architecture: (I cannot attach the file from links nor the logs) I am not sure if this is a request for enhancement.
Created attachment 150077 [details, diff] Fix (diff/patch) for the problem I have tested it on "linux-headers".
Created attachment 150079 [details] Console log of emerge before the fix
Created attachment 150081 [details] Shorted console log of emerge
unless you can give me a realistic reason why x86_64-pc-linux-gnu is not acceptable to you, i'm not going to merge this
(In reply to comment #4) > unless you can give me a realistic reason why x86_64-pc-linux-gnu is not > acceptable to you, i'm not going to merge this 1) An AMD Athlon is NOT x86_64 Architecture - Its is 32-bit! 2) As I understand it, a "Target" as per gcc and so on, is of the format: <m/c-name> - <O/S-type> - <manufacturer> The last 2 form "-linux-gnu". <m/c-name> for IBM PC's and Clones therof is <cpu-type>-pc. <cpu-type> is as defined on the GCC web-page (online documentation) under the -mtune and thereby the -march parameter values. From the above "athlon-pc-linux-gnu" should be a valid "Target" ie: CHOST CTARGET or CBUILD value. I agree that the many names for ARCH="x86" is very confusing. I was of the intent to normalise the naming convention for my machines - starting with amd/athlon (GRP/CHOST). I have other machines amongst which Pentium-3 and -2 as well as amd64/Opteron. At the end of the day it has more to do with completeness than anything else.
i686-... or x86_64-..., the answer is the same ... you havent given a valid reason for such a non-standard target