Another way to salvage Annex K and its constraint handler

[Rich Felker]

On Fri, Nov 20, 2015 at 04:17:22PM +0100, Florian Weimer wrote:
> While looking at __dso_handle and implementing a no-fail wrapper for
> getrandom, it occurred to me that we can implement Annex K in a useful
> fashion, with the constraint handler:
> 
> We put the variable that holds the constraint handler address into
> libc_nonshared.a (maybe even as a weak symbol), and pass the address of
> the variable to the actual Annex K implementation functions (with the
> help of inline wrappers, probably).  This way, each DSO can install the
> constraint handler it wants.  This would remove the need for a
> per-thread constraint handler (which would require constant switching of
> the handler at library boundaries) because libraries know when they
> change constraint handlers and implement suitable locking.
> 
> DSOs are out of the scope of C standard, so no standards action is
> needed here.

dlopen is outside the scope, but the behavior of a conforming program
that just happens to be dynamic-linked rather than static-linked is
not outside the scope. You cannot just ignore the constraint handler
set by one function because it happened to be performed from code in a
different DSO. What DSO the code lies in is an implementation detail.
And of course such a "solution" does nothing for static linking
anyway.

Rich