Re: What's the deal with building intermediate .o vs .a files?

[Graham Stoney]

> Graham Stoney wrote:
> > Is there some other hack I could
> > use to get "ar" to work with initcall, like treating it as a C++ constructor
> > or using KEEP in the .lds file or something?

Jeff Garzik writes:
> If there is an easy way to hack the ld script such that ar pulls in
> initcall [other similar special sections], that would cut down on some
> errors, I would think.

I'm currently experimenting with 2.2.x, which doesn't have initcall. I haven't
seen any problem with other lost sections yet.

Looking at initcall the 2.3.x kernel, I can't see what would make it disappear.
Can anyone enlighten me on this please?

> The general thinking seems to be "your code shouldn't contain dead code
> anyway", so getting ar to work with initcall has a fairly low priority
> IMHO

Sure, but it's real hard to do for any arbitrary combination of CONFIG_*
options, where you end up with a maze of #ifdefs. It's also error prone, and
not everyone cares enough to bother. I'm trying to go one better than ar, and
get the linker to do it for us. It's non-trivial, but I'm getting there
(slowly)...

Regards,
Graham