1 |
Hi François,
|
2 |
|
3 |
François Bissey <francois.bissey@×××××××××××××.nz> writes:
|
4 |
|
5 |
> I now have a working RAP environment on ppc64. |
6 |
> So here is a bit of recipe to achieve this: |
7 |
> * There is no current ppc64 profile, so one has to be |
8 |
> created during stage 1. |
9 |
> I copied the ppc64 from prefix/linux to prefix/linux-standalone. |
10 |
> Replace make.default inside prefix/linux-standalone/ppc64 |
11 |
> by the one in prefix/linux-standalone/amd64. |
12 |
> Copy the "legacy" folder from prefix/linux-standalone/amd64 |
13 |
> to prefix/linux-standalone/ppc64. |
14 |
|
15 |
Thanks!
|
16 |
|
17 |
> * Make sure help2man is installed on the host system (for bison). |
18 |
|
19 |
I don't remember if the bare minimal OpenSUSE 42.1 needed this.
|
20 |
|
21 |
> * I would like the following patch to be considered |
22 |
> for the bootstrap script |
23 |
> --- bootstrap-rap.sh.orig 2017-01-24 12:20:27.304355158 +1300 |
24 |
> |
25 |
> |
26 |
> +++ bootstrap-rap.sh 2017-01-24 12:34:30.325185905 +1300 |
27 |
> |
28 |
> |
29 |
> @@ -1382,8 +1382,8 @@ |
30 |
> |
31 |
> |
32 |
> sys-libs/ncurses |
33 |
> |
34 |
> |
35 |
> sys-libs/readline |
36 |
> |
37 |
> |
38 |
> app-shells/bash |
39 |
> |
40 |
> |
41 |
> - sys-apps/sed |
42 |
> app-arch/xz-utils |
43 |
> + sys-apps/sed |
44 |
|
45 |
> The latest version of sed is compressed with xz which |
46 |
> means the bootstrap will fail if xz-utils is not installed |
47 |
> in /bin or /usr/bin. |
48 |
> Newer distro probably have it but not older ones. |
49 |
> So we should move xz-utils before sed. |
50 |
|
51 |
Which tarball (URL) did it use?
|
52 |
|
53 |
> + PERL_NOT_INSTALLED=0 |
54 |
> |
55 |
> |
56 |
> if [[ ! -x "${ROOT}"/usr/bin/perl ]]; then |
57 |
> |
58 |
> |
59 |
> # trick "perl -V:apiversion" check of |
60 |
> glibc-2.19. |
61 |
> |
62 |
> echo -e "#!${ROOT}/bin/sh\necho |
63 |
> 'apiversion=9999'" > "${ROOT}"/usr/bin/perl |
64 |
> |
65 |
> chmod +x "${ROOT}"/usr/bin/perl |
66 |
> |
67 |
> |
68 |
> + PERL_NOT_INSTALLED=1 |
69 |
> |
70 |
> - rm "${ROOT}"/usr/bin/perl |
71 |
> |
72 |
> |
73 |
> + # Do not mindlessly remove perl. Only if it wasnt |
74 |
> present in the first place. |
75 |
> |
76 |
> + [[ ${PERL_NOT_INSTALLED} == 1 ]] && rm |
77 |
> "${ROOT}"/usr/bin/perl |
78 |
> |
79 |
> else |
80 |
|
81 |
> The other part of the patch is specific to RAP and is |
82 |
> vexing. RAP adds a fake perl executable before building |
83 |
> glibc in case perl is not already there - and remove it, |
84 |
> unconditionally. So you are here you had a failure in stage3 |
85 |
> after glibc has been merged and then perl has been merged. |
86 |
> You fixed whatever needs fixing and restart the bootstrap |
87 |
> and it removes your perl binary. And then you spend some |
88 |
> time wondering why autoconf failed while emerging something. |
89 |
> So I added a guard - only remove the perl executable if it is |
90 |
> the fake one. I guess that could be improved by actually |
91 |
> checking the file itself for the content of the fake perl. |
92 |
|
93 |
Sorry for this, I added a guard
|
94 |
|
95 |
https://gitweb.gentoo.org/repo/proj/prefix.git/commit/?id=db54c48edfb3d7c5c482326c8c57d3823941f456
|
96 |
|
97 |
> * As per another post the current bits to set the RAP glibc |
98 |
> to be used by gcc are not sufficient for ppc64. My profile.bashrc |
99 |
> has the following |
100 |
> if [[ ${CATEGORY}/${PN} == sys-devel/gcc && ${EBUILD_PHASE} == |
101 |
> configure ]]; then |
102 |
> cd "${S}" |
103 |
> einfo "Prefixifying dynamic linkers..." |
104 |
> for h in gcc/config/*/*linux*.h; do |
105 |
> ebegin " Updating $h" |
106 |
> sed -i -r "/_DYNAMIC_LINKER/s,([\":])(/lib),\1${EPREFIX}\2,g" $h |
107 |
> eend $? |
108 |
> done |
109 |
> |
110 |
> |
111 |
> The only change being to the sed line. |
112 |
|
113 |
Nice.
|
114 |
|
115 |
I am wondering if it is feasible to grant me access to a ppc64 box with
|
116 |
SLES 11.1 to test out these patch before committing to the
|
117 |
repositories.
|
118 |
|
119 |
Yours,
|
120 |
Benda |