Gentoo Archives: gentoo-user

From: "Timothy A. Holmes" <tholmes@×××××××××.net>
To: gentoo-user@l.g.o
Subject: RE: [gentoo-user] changing CHOST
Date: Sun, 10 Sep 2006 22:28:43
Message-Id: 17CD9CE4C0FA574A8B29EF02D49B385D05BC1A@srvexch-01.mcaschool.local
1 -----Original Message-----
2 From: richard.j.fish@×××××.com on behalf of Richard Fish
3 Sent: Sun 9/10/2006 2:30 PM
4 To: gentoo-user@l.g.o
5 Subject: Re: [gentoo-user] changing CHOST
6
7 On 9/10/06, Timothy A. Holmes <tholmes@×××××××××.net> wrote:
8 > Richard:
9 >
10 > I got this one - thanks, the problem is that im getting conflicting
11 > advice from multiple sources.
12
13 I think the problem with this is (and I don't mean to offend when I
14 say this...) that the people who would be able to figure out how to
15 successfully change CHOST on a live system have never have to do so,
16 because they set it right from the start.
17
18 So my advice is based on:
19
20 http://article.gmane.org/gmane.linux.gentoo.user/169260
21
22 As well as forums threads such as:
23
24 http://forums.gentoo.org/viewtopic-t-386633.html
25
26 (BTW, one has to be a bit careful on the forums, as some of those
27 threads have answers from people who recommend merging system and
28 world twice each....which is pretty much useless.)
29
30 > You cant do that no matter what
31
32 Pretty sure this isn't true. But it would be wrong to suggest that it
33 is an easy thing to change, or say that there is some method that
34 guarantees you won't end up booting from your live CD to repair
35 things.
36
37 > To
38 > You don't want to do that
39
40 This one seems entirely up to you. At this point in Gentoo, you
41 either have to change CHOST, or add ">=sys-libs/glibc-2.4" to
42 /etc/portage/package.mask, since 2.4 is nptl only, and that requires
43 better than i386.
44
45 > To
46 > It wont do any good
47
48 If this was the entire answer, it is simply clueless. Using the CHOST
49 that matches your processor lets gcc use more effecient instructions
50 for newer processors. This is what makes nptl so more efficient than
51 linuxthreads, because it uses processor instructions specifically
52 designed for multi-CPU synchronization.
53
54 Now whether the improvements are worthwhile or not is a subjective
55 thing, and one could argue that it isn't worth the effort. This goes
56 back to the previous point.
57
58 > To
59 > Bootstrap wont run
60
61 Sounds like a bug.
62
63 > To
64 > No need for bootstrap, just change and do the emerges
65
66 Well I would have thought so too, but that hasn't worked for some
67 people. Again, there is no definitive guide on changing CHOST. The
68 safest option is to boot from a livecd and re-install using the new
69 CHOST.
70
71 But if you are willing to go that far anyway, it can't hurt to try the
72 bootstrap.sh ; emerge -e system ; emerge -e world sequence.
73 Basically, if you make it through the emerge -e system part, you have
74 a sane base system and anything else that breaks indicates a problem
75 with the change in profile or gcc versions, not the change in CHOST.
76
77 BTW, Darren's answer on this thread seems incorrect to me. Changing
78 CHOST is a pretty significant thing to tweak, certainly as significant
79 as changing gcc versions, and you really should re-merge *everything*
80 to make sure your something doesn't wind up broken.
81
82 > Im a bit confused to be honest
83
84 That's ok. It isn't an easy question to answer unfortunately.
85
86 -Richard
87 --
88 gentoo-user@g.o mailing list
89
90
91 Richard:
92
93 thanks a great deal for your answer -- it is extremely helpful.
94
95 I think i will give it a shot, the system in question right now is only my rsync mirror -- it as additional tasks planned to be installed, but that hasn't happened yet. if i can get it to run, then it is cool, but if not, i havent lost a great deal. I do however have a couple other production boxes that I REALLY dont want to have to rebuild that might (i havent checked yet) be built using the same stage 3 -- they are running ok, speeds and processor loads look ok, so if this breaks this box, im not going to attempt it on the other ones.
96
97 I'll let you know how it goes
98
99 TIM

Attachments

File name MIME type
winmail.dat application/ms-tnef