On Thu, Mar 02, 2006 at 06:53:16PM +0100, Aurelien Jarno wrote:
[snip]
Other architectures have a simple separation between 32 and 64 bit. For
Mips there's often some confusion what n32 qualifies for (embedded
people often call it 64 bit without further qualification). That's why
I recommended to use the ABI names instead. This also avoids the heavily
overloaded "mips32" and "mips64" terms.
You mean for the libc provided packages or in general?
In general.
If you mean something like libn32gcc1, libn64gcc1 or libn32z1, libn64z1,
just be aware that it makes patches to bi/tri-arch packages a lot
bigger. If using 32 or 64, it usually means just adding mips and mipsel
to a variable.
And then people have to remember that lib32gcc1 provides _not_ the usual
32bit ABI (and doesn't even work on a 32bit CPU, or a 32bit kernel
running on a 64bit capable CPU). This usually goes wrong.