1 |
On 24/08/2015 15:17, Alan Mackenzie wrote: |
2 |
> Hello, Harry, |
3 |
> |
4 |
> Long time, no see! |
5 |
> |
6 |
> On Sun, Aug 23, 2015 at 10:19:42PM -0400, Harry Putnam wrote: |
7 |
>> My gentoo OS is running on Openindiana (solaris) inside oracle's vbox. |
8 |
> |
9 |
>> It's been left setting for at least 4-5 months maybe a couple more. |
10 |
> |
11 |
>> After eix-sync, attempting an `emerge vuND world' comes up with so |
12 |
>> many blocks, use flag changes and a variety of other bad news in |
13 |
>> such proliferation... I'm thinking better to install from scratch with |
14 |
>> latest ISO. |
15 |
> |
16 |
> I know the feeling. When the same thing happened to me (my system got |
17 |
> into a mess because I was running XFCE which was too dependent on |
18 |
> gnome-2; when gnome-3 became stable, the demons of hades were let |
19 |
> loose, and I tried a couple of times, half-heartedly, to update), I told |
20 |
> myself to stop and think. |
21 |
> |
22 |
> I could spend days fighting with use flags and conflicts, or I could |
23 |
> spend a few days reinstalling. In the end I reinstalled (using the old |
24 |
> system (rather than an ISO image) to do the initial stages. It took me |
25 |
> about a week, compared with about a month the first time I seriously |
26 |
> installed Gentoo. "Installation" here means getting everything up and |
27 |
> running, including X with destop manager, printing, sound, email server |
28 |
> and client, .... |
29 |
|
30 |
|
31 |
At first glance it does look like maybe a reinstall would be better. But |
32 |
in this case, that's not true. Looking over the list of packages to be |
33 |
updated, there are 3 general classes of things: |
34 |
|
35 |
1. Regular updates |
36 |
2. A whole whack of rebuilds |
37 |
3. A perl upgrade from 5.20 to 5.22 |
38 |
|
39 |
#1 is routine. Press enter, and make does it's thing |
40 |
#2 looks scary, but in the old days we'd have to do the updates then let |
41 |
revdep-rebuild catch the inconsistencies, and rebuild those. Modern |
42 |
portage has some magic code to fold everything into the main emerge |
43 |
world step. So just press Enter and make does it's thing |
44 |
#3 can be very confusing. With 5.22, upstream moved many Perl packages |
45 |
into the core Perl codebase, so all such installed packages (and all of |
46 |
us have many of them) need to uninstall perl-core/<package> and replace |
47 |
it with virtual/perl-<package>. Portage normally deals with this |
48 |
transparently, but the output can be a little too verbose sometimes, and |
49 |
takes some decent brainpower to figure out what is really going on. |
50 |
|
51 |
There's also some blockers in that list ([blocks b ]), but they are all |
52 |
soft (lower case b) so portage should take those in it's stride and just |
53 |
fix it with no intervention. |
54 |
|
55 |
|
56 |
Harry's real problem as many have noted is that his gcc config is not |
57 |
valid. But, portage can't tell him that. It started the merge, and |
58 |
handed control over to the next app, which portage can't make sense of |
59 |
as it all happened outside portage's control. Result: a wall of text on |
60 |
the screen, right after the wall of text of 195 things to be rebuilt and |
61 |
a huge list of stuff causing other stuff to be needed to be rebuilt! |
62 |
|
63 |
I suppose those ebuilds that are sensitive to gcc versions could have a |
64 |
check built in to check the version before starting and then print a |
65 |
sane error message that portage CAN control and make sense of. |
66 |
|
67 |
Either way, "gcc-config -l" is what Harry needs to run first, and make |
68 |
his default compiler 4.7 or later. With that out of the way, emerge |
69 |
should proceed normally apart from taking a while to get through it. He |
70 |
might have perl issues afterwards and need to run perl-cleaner. |
71 |
|
72 |
The real problem is how do I know all this, and Harry did spot it? Well, |
73 |
building a source distro comes in at a much lower level than a binary |
74 |
one, and it does take a good large dose of experience, knowledge and |
75 |
plain old luck to figure out what is really going on. These days portage |
76 |
is very good at doing the right thing (it still sucks at a human level |
77 |
in it's output...) and today Harry just got really unlucky. |
78 |
|
79 |
So yeah, reinstall is probably not the better option here. |
80 |
|
81 |
|
82 |
-- |
83 |
Alan McKinnon |
84 |
alan.mckinnon@×××××.com |