Re: -fno-inline-functions vs glibc's initfini

[Roland McGrath <roland at hack dot frob dot com>]

> Want me to prepare a s/-fno-inline-functions/-fno-inline/ patch?

My reading was that actually we would want both of those switches (the
first to avoid inlining as an implicit optimization, the second to avoid
inlining of functions declared inline).  But whatever the exact detail,
yes, please send a patch that does the right thing.

> Maybe rather than using the generic version, we should have a Makefile
> rule that generates the machine-dependent .s files for developers'
> perusal in creating the machine-specific asm sources.

That certainly wouldn't hurt.  I guess we could start with that to make it
easy to produce the assembly code, and then let each arch add one piecemeal
before taking away the C option.  Once we get to every arch having
assembly, it probably makes sense to stop doing it as a single file with
magic boundary markers to be sliced, and just have each arch provide
crt[in].S files directly.


Thanks,
Roland