1 |
Hi, |
2 |
|
3 |
I've again got a platform support question: |
4 |
|
5 |
Has the current base toolchain for PPC |
6 |
(binutils-apple-4.2/gcc-4.2.1_p5666-r2) ever been tested on OS X 10.4 PPC? |
7 |
|
8 |
I ran into serious trouble trying to bootstrap my old iBook. |
9 |
binutils-apple-4.2 doesn't compile. After fixing that, it can't link |
10 |
gcc-apple-4.2.1's cc1objplus because the entry __start and main() are |
11 |
more that 16MB apart. This got me on to crt1.o from csu which needs to |
12 |
be compiled with -mlong-branch on 10.4 which is isn't. After fixing that |
13 |
I now get an error that "half word text relocs aren't supported in |
14 |
__start for architecture ppc". I don't even know where to start. :-/ |
15 |
|
16 |
Trying all this with binutils-apple-3.2, which do compile, doesn't |
17 |
change things either. |
18 |
|
19 |
So my guess is that when prefix was working on 10.4 PPC, much more stuff |
20 |
was used from the system directly, such as crt1.o. Perhaps even |
21 |
gcc-4.0.1 was still in use. |
22 |
|
23 |
Is someone currently actually using prefix on OS X 10.4 PPC and can help |
24 |
me out? |
25 |
|
26 |
Rationale: 10.4 is the ultimate benchmark for self-containedness of |
27 |
prefix because loads of stuff the tools in the prefix need just don't |
28 |
exist in the system yet. It got me on to clang always using the system's |
29 |
libstdc++ for example because that produces warnings because of a |
30 |
mismatching libgcc_s from the prefix. |
31 |
-- |
32 |
Thanks! |
33 |
Michael |