Emerge vanilla-sources was failing the function universal_unpack in kernel-2.eclass. I changed the line [[ ${OKV} != ${KV_FULL} ]] && mv linux-${OKV} linux-${KV_FULL} \ || die "Unable to move source tree to ${KV_FULL}." to if [[ ${OKV} != ${KV_FULL} ]]; then if mv linux-${OKV} linux-${KV_FULL} ; then die "Unable to move source tree to ${KV_FULL}." fi fi For some reason the first one was thinking that ${OKV} and ${KV_FULL}, both set to 2.4.28 were different and was trying to mv linux-2.4.28 linux-2.4.28 which was failing. After changing the line to the slightly less slick if statement it appeared to work fine. Output of emerge below. mahler:/home/jorussel# emerge -v vanilla-sources Calculating dependencies ...done! >>> emerge (1 of 1) sys-kernel/vanilla-sources-2.4.28 to / >>> md5 src_uri ;-) linux-2.4.28.tar.bz2 >>> Preparing to unpack ... >>> Unpacking source... >>> Unpacking linux-2.4.28.tar.bz2 to /var/tmp/portage/vanilla-sources-2.4.28/work !!! ERROR: sys-kernel/vanilla-sources-2.4.28 failed. !!! Function universal_unpack, Line 163, Exitcode 1 !!! Unable to move source tree to 2.4.28. !!! If you need support, post the topmost build error, NOT this status message. Reproducible: Always Steps to Reproduce: 1.emerge vanilla sources-2.4.28 2.emerge fails 3. Actual Results: emerge failes in universal_unpack function Expected Results: universal_unpack should not try to move the directory linux-2.4.28 to linux-2.4.28.
Can we have a working kernel-2.eclass in portage? Is there testing for eclasses before putting them in Portage? We rely on a working implementation of Portage here, I don't understand why bugs like this make their way in the tree. A way to keep the shorcuts in could be: [[ ${OKV} != ${KV_FULL} ]] && ( mv linux-${OKV} linux-${KV_FULL} \ || die "Unable to move source tree to ${KV_FULL}." ) With the current line it doesn't even try to mv. Since ${OKV} != ${KV_FULL} in this case, the result of && is automatically false, and so it dies. Can we have a fix rapidly?
Not so sound pretentious, but I think that the if statement is far more readable than all the && || &|&&| magic. I know its not as cool, but it'll make someone else's life easier later.
Looks like this has been fixed, please reopen this bug if the issue still occurs. RCS file: /var/cvsroot/gentoo-x86/eclass/kernel-2.eclass,v head: 1.98 ---------------------------- revision 1.98 date: 2005/02/09 21:07:14; author: hollow; state: Exp; lines: +4 -2 fixed universal_unpack ----------------------------
