[PATCH 09/12] RISC-V: Atomic and Locking Routines

[Joseph Myers]

On Fri, 23 Jun 2017, Torvald Riegel wrote:

> Thanks for providing some background.  However, elsewhere in the thread
> for the whole patch set, you seemed to say that you'll only officially
> support target configurations that do have atomics (IIRC, that was a
> discussion with Joseph about the target triples).  These officially
> supported ones are the configurations we'll test with
> build-many-glibcs.py.  I think that it would be better to only have code
> in master that will be compiled by some build through
> build-many-glibcs.py (ie, that there's no dead code from that build
> coverage perspective).

I'm pretty sure lots of existing glibc architectures have code that's not 
covered by build-many-glibcs.py at present, even if there are 
configurations that could be added (e.g. building for another CPU variant) 
that would cover building that code.  That doesn't mean it's a good thing 
- more configurations to cover such code would be a good idea - but the 
present minimum for build-many-glibcs.py coverage is that at least each 
different ABI supported by glibcs is covered.

-- 
Joseph S. Myers
joseph@codesourcery.com