The behaviour of mirrorselect when ipv6 is set in USE-flags should imho be the following: <snip> --- /usr/sbin/mirrorselect 2004-03-23 00:34:46.000000000 +0100 +++ mirrorselect 2004-04-16 11:12:28.174255084 +0200 @@ -362,8 +362,8 @@ strip_ipv6 elif [ -z ${DEEP} ] then - echo "Netselect is currently unable to handle ipv6, changing to Deep Test" - DEEP="1" + echo "Netselect is currently unable to handle ipv6, dropping ipv6 hosts" + strip_ipv6 fi [ -n "${DEEP}" ] && deeptest || automatic write_config </snip> Cheers, Tobias
fixing summary
Also mirrorselect will write "Netselect is currently unable to handle ipv6, changing to Deep Test" to /etc/make.conf with out commenting it out, this causes env-update to fail later in the process. Message should not be written to make.conf or should be commented out.
Or it should simply be echoed to stderror instead of to stdout. Which would fix this problem and still give the error message.
Created attachment 32223 [details, diff] suggested patch to mirrorselect to stop output of netselect error if -o
Just to make clear. My initial suggest was about not changing to deep-test when ipv6 is used. The output did not bother me.
My suggested patch is to fix the problem stated by Reinaldo (Comment #2). Not sure if this is the right bug report to post to regarding this problem. This part of the bug is reported in bug 52264 (http://bugs.gentoo.org/show_bug.cgi?id=52264), but it has been assigned to the docs team. I'm not sure why the mirrorselect script changes to a deep test rather than dropping the ipv6 hosts, as Tobias suggests. But whether the behaviour of the script is changed as per Tobias's suggestion or not, it will still need to either not ouptut to stdout if the -o option is passed to it, or the line should simply be commented, i.e. line 365: echo "# Netselect is currently unable to handle ipv6, ....." so that it doesn't cause env-update to fail. It will also fix bug 52246 as the documentation won't need to be changed since there wouldn't be any need to edit make.conf :-) Cheez, Mahmoud
I mean it will fix bug 52264, not 52246 :-o
I was about to file the same bug, Tobias. :-) I agree: Please do not go to deep testing on IPv6-enabled hosts. It makes the utility almost unusable for me.
behaviour changed to drop ipv6 (or should I say, those known hosts) regardless. fixed to echo to stderr. should be in portage with mirrorselect-0.85. thanks