Re: Solving the recursive memcpy/memset/etc. issue

[Rich Felker <dalias@aerifal.cx>]

On Wed, Jul 31, 2013 at 08:49:40PM -0400, Rich Felker wrote:
> OK, so now that it's hit us for real, what should we do about GCC
> generating code for memcpy, memset, etc. which might contain infinite
> recursion? Aside from the ARM issue (which was separate), we know the
> option causing this bad code generation, and it can be disabled via
> -fno-tree-loop-distribute-patterns. However, if GCC policy is that
> they consider the compiler entitled to generate calls to
> memcpy/memset/memmove/memcmp whenever it wants, then we're just going
> to be playing whack-a-mole.

For now, I solved it with just this option, and also added some more
asm for x86[_64].

> The only fully viable option I see is replacing the code for these
> functions with code that uses volatile objects so as to make
> optimization utterly impossible. This will of course make them
> incredibly slow, but at least we would have safe, working C code, and
> we could add asm for each supported arch.

I'd actually like to experiment with some ideas for volatile that
don't make the C slow. If the compiler really honors volatile, it
seems like you can write C code in such a way where the compiler has
to implement it pretty close to the abstract machine -- but you can be
selective about it, so as to allow the compiler some freedom.

> An alternative might be to test the compiler in configure to determine
> if, with the selected CFLAGS, it generates recursive code for these
> functions, and if so, defining a macro that causes musl to revert to
> the volatile code.

This is not in the workaround I committed, but could be added..

Rich