On Thu, Jun 21, 2007 at 11:47:08AM -0500, Mark Hatle wrote:
> I don't think that the "j" makes a lot of sense. I've never seen any
> code that actually makes use of the 'j'.
A Linux 2.6 kernel running on an ARM926EJ-S, an ARM1026EJ-S or a
Marvell Feroceon core will report 'armv5tej' in uname -r.
See linux-2.6.*/arch/arm/proc-arm{926,1026}.S.
> Also do you have any provisions for specifying "vfp" support? If
> not, you may want to consider that as another permutation (if you
> are trying to be "complete".)
Well, the current set only reflects the various `uname -r` style ARM
archs that exist, and the presence of VFP is typically not encoded in
the arch name.
Of course, we could come up with our own name for it ('armv5tejvl'?
'foo'?), but then rpm (specifically, lib/rpmrc.c:defaultMachine())
would also need some hacking not to use the literal output of
SYS_uname as the arch name, but take HWCAP_VFP into account.
