Re: [tree-ssa][rfc] plans for nested functions

[Linus Torvalds <torvalds at osdl dot org>]

Richard Henderson wrote:
>
> This is stuff that was running around my head this morning.
> I mostly wanted to write this down so that I don't forget...
> 
> The problem I'm trying to solve is that, at present, we have
> ordering issues wrt compilation of nested functions.
> 
> (1) We depend heavily on rtl-level state in order to compile
>     nested functions.

Why do you want to continue to support nested functions in the first
place? They are an abomination, a gcc-specific extension, and nobody
sane uses them (yeah, yeah, libobj, but I said _sane_).

Some ideas are just bad. Nested functions was one such idea. The
implementation used to do bad things (icache flushes at run-time,
for christ sake!).

Maybe you could support a specific subcase of nested functions where
they are always inlined, and basically do some slightly less nasty
tree-based inliner thing rather than any real nesting. And document
the horror as being a bad idea and only supported for legacy gcc
reasons?

Are there really any users that do not deserve to either be fixed
or just shot in the head?

                Linus