[Clfs-dev] EGLIBC x86 is special?

[Andrew Bradford]

In trying to learn how to convert the embedded book from uClibc to
EGLIBC I noticed that the first building of EGLIBC in the main book uses
slightly different configure options for x86 vs the other archs.

For x86 32 bit and x86 32 bit portion of multilib but not for any other
arch including x86_64 there's the addition of:

CFLAGS="-march=$(cut -d- -f1 <<< $CLFS_TARGET32) -mtune=generic -g -O2"

or similar passed to configure.

This looks like a hold-over from glibc, as shown by commit SHA:
093b0e8b160919f5e1cf7aac272d710a68302a3c where that line is still
present while changing from glibc to EGLIBC back in 2009.

Is that line really needed?

Shouldn't the fact that the target is being set properly be enough to
set -march= and do we really need to -mtune= or include debug symbols or
set optimizations to O2?  I believe most of these are the defaults.  But
regardless, if they're not, why does only 32 bit x86 get this treatment?

Thanks,
Andrew