1 |
"Christopher E" <sensory.access@×××××.com> posted |
2 |
184c54640605200819k795c6e4dma6bf1fc302f66c2e@××××××××××.com, excerpted |
3 |
below, on Sat, 20 May 2006 11:19:18 -0400: |
4 |
|
5 |
> Hello Ducan, |
6 |
> |
7 |
>> I've only done the initial stage-1 setup, and now they recommend |
8 |
>> starting from stage three and just doing an emerge --emptytree, so I |
9 |
>> don't have much of use to say in that or the other areas. |
10 |
> |
11 |
> Do you have any stuff I can read about using a stage three and doing a |
12 |
> emerge --emptytree, and why one would do such? |
13 |
|
14 |
The newer editions of the handbook cover it in some detail, but I was |
15 |
lurking on the dev list when it was discussed there, so I can give a bit |
16 |
more developer side viewpoint. Boyd SS Jr. also has some very good |
17 |
points, so grab them too. (That's what's great about lists/groups, often |
18 |
you get several posts, each from slightly different viewpoints, and get to |
19 |
pick and choose the explanations and try the solutions first that fit |
20 |
yourr style. =8^) |
21 |
|
22 |
Basically what's happened is that with all the packages and platforms and |
23 |
all the customization possible with Gentoo now, it's simply no longer |
24 |
possible to test all the permutations and corner-cases of building from a |
25 |
stage-1. Too often, some of those corner cases fail to work, and people |
26 |
trying to install Gentoo are left frustrated when they run into first one |
27 |
problem and then another. |
28 |
|
29 |
The stage-3 approach, which is what the handbook strongly recommends now, |
30 |
starts with far more of the system initially compiled. You then do an |
31 |
emerge --empty-tree system once you've set up your CFLAGS and the like the |
32 |
way you want them, and portage recompiles everything the way you want it. |
33 |
Done right, it's a far easier way to end up with the same customized |
34 |
system you'd have otherwise, and since the packages in the stage-three are |
35 |
all tested to work as they exist in the stage-three, it's far harder to |
36 |
break and there are far fewer corner cases for folks to run into. |
37 |
|
38 |
There are a few drawbacks, however. As BSSJ mentions, it's a bit more |
39 |
difficult to change CHOST, among other things. However, with amd64 we are |
40 |
lucky as there's really not much reason for us /to/ change CHOST. The |
41 |
other one that's big here, is that I not only started from stage-one, but |
42 |
due to those memory issues I mentioned in the reply to BSSJ, I actually |
43 |
had to take apart bootstrap.sh and execute each step manually, one at a |
44 |
time. As a result, I know rather more about what makes a Linux system tick |
45 |
than I would have otherwise. Folks that do a stage-three simply aren't |
46 |
going to be forced to get as up close and personal with their systems as |
47 |
those choosing a stage-one and persevering thru all the corner-cases to |
48 |
find solutions. It'll be far faster and far easier, but they'll not know |
49 |
as much about their system when they are done, and in some ways, that's a |
50 |
shame. It actually took me three months to get the system up and running |
51 |
(I was dual booting Mandrake at the time, so it wasn't as if the system |
52 |
was entirely dead for three months), due to the hardware and some software |
53 |
problems I ran into, back on 2004.0 and 2004.1 when I first installed |
54 |
Linux, but I'm glad I did it that way, as I know FAR more about Linux and |
55 |
Gentoo than I would have otherwise. (As well, since it took me that long, |
56 |
and all the while I was in this group/list, the user group, and lurking in |
57 |
the devel group, I learned all the stupid mistakes folks make before I |
58 |
got to them and was able to steer clear. Stuff like the guy that just ran |
59 |
emerge --prune without checking to see what it was doing, first, etc.) |
60 |
|
61 |
>> Specific: Check that those video card USE flags are all USE flags, not |
62 |
>> VIDEO_CARDS flags (that is, USE_EXPAND, not USE). sis, for instance, |
63 |
>> doesn't appear to be a USE flag (euse -i sis doesn't list it for either |
64 |
>> local or global), altho i8x0 is a local use flag for a couple packages. |
65 |
> |
66 |
> I in this case am not using the sis or what not. I am going to just use |
67 |
> the video card that I am wanting to have support for under the VIDEO_CARDS |
68 |
> flags like radeon, vesa etc when I understand more about telling it not to |
69 |
> install support for others then I will add this then. |
70 |
|
71 |
BSSJ covered this. All I was saying is be sure to use --pretend, and get |
72 |
the right stuff in the right place, USE vs VIDEO_CARDS, or you'll not end |
73 |
up with the system you wanted. However, it looks like you got that |
74 |
covered. |
75 |
|
76 |
>> General: Since you are starting over, it's a good time to consider one |
77 |
>> of my favorite features, FEATURES=buildpkg |
78 |
> Can you tell me more about this and is this go in the make.conf file? so |
79 |
> does this mean the system runs from bin files instead of what every the |
80 |
> other way is? |
81 |
|
82 |
BSSJ covered this, too. I'll just add a pointer back to the handbook. In |
83 |
particular, this is mentioned (but unfortunately not in the detail I'd |
84 |
prefer if I wrote it) along with a bunch of other useful info, in the |
85 |
Working with Portage and Working with Gentoo sections of the handbook. |
86 |
DON'T JUST READ THE INSTALL SECTION AND STOP, OR YOU'LL MISS SOME OF THE |
87 |
BEST PARTS OF THE HANDBOOK! Those Working with... parts are extremely |
88 |
useful for just that... continuing to work with your system as the weeks |
89 |
and months roll by, long after you finish installing. Again, since it |
90 |
took me so long to get a working system up and running, I had plenty of |
91 |
time to read up on those things, and read thru them several times, and |
92 |
from what I've seen on the groups, I'm more effective as a Gentoo sysadmin |
93 |
than most, because I know the tools and how they work. |
94 |
|
95 |
Very briefly since BSSJ covered it and it's in the documentation too, |
96 |
after the compile and install to the fake image in $PORTAGE_TMPDIR are |
97 |
done, portage normally does a qmerge step, that moves everything from the |
98 |
fake install image to the live system all in one step. FEATURES=buildpkg |
99 |
changes that so portage creates its own binary package archive instead, |
100 |
then unpacks the archive and installs from it, thereby testing the archive |
101 |
as it installs. Again, having this archive can be very handy if an upgrade |
102 |
goes wrong and you need to remerge an older version again. Since you had |
103 |
it merged before, you should have the binpkg still around, and can tell |
104 |
portage to use it rather than having to go thru and recompile it again, as |
105 |
you would if you'd removed it to install the upgrade then decided to roll |
106 |
back, but did NOT have the binpkg hanging around. |
107 |
|
108 |
I meant to mention the first time but forgot. The binpkg archive is |
109 |
normally kept in your portage tree, and will take 2-4 gigs more space than |
110 |
if you didn't have it. |
111 |
|
112 |
As for ccache and confcache, ccache is dealt with in the handbook fairly |
113 |
well, so no need to cover it here. confcache is newer, new to still ~arch |
114 |
portage 2.1-rcs, but there's very little to configure there, besides |
115 |
turning on the feature and emerging confcache. I believe it's covered at |
116 |
least in make.conf.example and the make.conf manpage, assuming you have |
117 |
~arch portage of course. However, I don't think it's in the handbook yet. |
118 |
The only thing you have to worry about with it is that very occasionally |
119 |
something might not compile right with it, but if you are running ~arch |
120 |
portage and gcc, you should be prepared to deal with that anyway. It's |
121 |
easy enough to toggle FEATURES=-confcache if you are having trouble with |
122 |
something, just to ensure it's not confcache. |
123 |
|
124 |
|
125 |
|
126 |
-- |
127 |
Duncan - List replies preferred. No HTML msgs. |
128 |
"Every nonfree program has a lord, a master -- |
129 |
and if you use the program, he is your master." Richard Stallman |
130 |
|
131 |
-- |
132 |
gentoo-amd64@g.o mailing list |