1 |
Quickie analysis... I followed the instructions in the HOWTO file at |
2 |
http://distfiles.gentoo.org/experimental/x86/musl/HOWTO probably the |
3 |
problem. It may have been possible to build git and layman against the |
4 |
standard portage tree, before activating layman when |
5 |
stage3-i686-musl-vanilla-20180304.tar.bz2 shipped. That's not the case |
6 |
now. The portage tree is constantly changing |
7 |
|
8 |
I followed the instructions, with a standard chroot from the minimal |
9 |
install ISO. Things break down at Step 3 of the HOWTO. Emerging |
10 |
layman+git wants to build 65 packages!!! This runs into into the |
11 |
problems noted in Step 5 (what happens if you don't activate layman |
12 |
before emerging world). Chicken+egg problem... I can't build layman+git |
13 |
without having functioning layman+git to set up the overlay first. |
14 |
Given that layman and git are crucial to installing the libmusl distro, |
15 |
please ship the stage3 tarball with a working layman+git. It may have |
16 |
worked when the stage3 tarball was released, but the portage tree is |
17 |
constantly changing. |
18 |
|
19 |
I admit that I might have goofed somewhere. But let's just say that |
20 |
I've done a lot of Gentoo installs, going back to the days of "stage 1", |
21 |
and I think I've followed the HOWTO properly. For diagnostics, see the |
22 |
attachment. |
23 |
|
24 |
OUTPUT 1 is the first attempt to emerge layman+git. Portage says to add |
25 |
">=dev-libs/openssl-1.0.2n -bindist" into /etc/portage/package.use |
26 |
|
27 |
OUTPUT 2 is what happens after adding the line to package.use. |
28 |
|
29 |
I did try unmasking+keywording as per OUTPUT 2. You do not want to |
30 |
see "OUTPUT 3". Even with --keep-going it eventually died and failed |
31 |
with multiple packages not building. |
32 |
|
33 |
A couple of additional comments... |
34 |
|
35 |
1) It would be nice to have app-portage/cpuid2cpuflags shipped as part |
36 |
of this (and all) stage3 tarballs. Many installs require an |
37 |
"emerge -e @world" step, so lets have the cpu flags set right when it's |
38 |
run. |
39 |
|
40 |
2) "eselect news list" gives stuff going back to 2013! We're on GCC |
41 |
6.4.0 already, so why worry about notices for GCC 4 and 5? The first 9 |
42 |
items below can be dropped from the tarball. |
43 |
|
44 |
(chroot) livecd / # eselect news list |
45 |
News items: |
46 |
[1] N 2013-09-27 Separate /usr on Linux requires initramfs |
47 |
[2] N 2014-06-15 GCC 4.8.3 defaults to -fstack-protector |
48 |
[3] N 2014-10-22 Upgrading to musl 1.1.5 |
49 |
[4] N 2014-10-26 GCC 4.7 Introduced the New C++11 ABI |
50 |
[5] N 2015-02-02 New portage plug-in sync system |
51 |
[6] N 2015-07-25 Python 3.4 enabled by default |
52 |
[7] N 2015-08-13 OpenSSH 7.0 disables ssh-dss keys by default |
53 |
[8] N 2015-10-22 GCC 5 Defaults to the New C++11 ABI |
54 |
[9] N 2016-06-19 L10N USE_EXPAND variable replacing LINGUAS |
55 |
[10] N 2018-01-30 Portage rsync tree verification |
56 |
[11] N 2018-03-13 Portage rsync tree verification unstable |
57 |
|
58 |
-- |
59 |
Walter Dnes <waltdnes@××××××××.org> |