| 1 |
David Cummings wrote: |
| 2 |
> I have a couple of questions I hope someone can shed some light on. |
| 3 |
> First, if I did a stage1 install using -mabi=32, I am using a o32 |
| 4 |
> userland? |
| 5 |
|
| 6 |
Yes. |
| 7 |
|
| 8 |
> Second, if I wanted to move to a n32 userland, without |
| 9 |
> reinstalling from netboot, what would be the steps involved? A change |
| 10 |
> of profile, emerge system && emerge world? or emerge --newuse --deep |
| 11 |
> world all in itself. |
| 12 |
|
| 13 |
Currently, this would require a complete reinstall. In any case, n32 |
| 14 |
userland is really not ready for general consumption yet. There are |
| 15 |
known bugs that we haven't squashed, and that could cause serious pain |
| 16 |
to end users. |
| 17 |
|
| 18 |
> And thirdly, if I want to compile a 64-bit |
| 19 |
> binary, the gcc I am currently using, mips-unknown-linux-gnu, claims |
| 20 |
> it is unable to produce an executable. |
| 21 |
|
| 22 |
Of course it can't. It has no 64-bit library to link the binary |
| 23 |
against. However, you should still be able to build 64-bit object |
| 24 |
files, and other programs that don't need a libc (like the kernel). |
| 25 |
|
| 26 |
> Will the -mabi=n32 one be OK? |
| 27 |
|
| 28 |
If you do install n32 userland, your toolchain will be |
| 29 |
mips64-unknown-linux-gnu, and will produce n32 userland binaries just |
| 30 |
fine via -mabi=n32. However, then -mabi=32 and -mabi=64 won't work, |
| 31 |
since there are no o32 or n64 libs to link against. Makes sense, yes? |
| 32 |
|
| 33 |
-Steve |
| 34 |
-- |
| 35 |
gentoo-mips@g.o mailing list |
Archives