[cfe-dev] Workaround for bug 16070 ?

[Stephen Kelly]

Tim Northover wrote:

>> It has been two weeks. Tim any further input?
> 
> I started a reply earlier, but somehow deleted it, sorry. I think it's
> a tricky situation because Cortex-A8 isn't the minimal CPU that could
> reasonably have VFP. GCC, for example, defaults to some ARM9 with an
> FPU added (which has some claim to being the minimal possible
> combination that would make sense for gnueabihf).
> 
> We don't have to follow them, of course. If no-one is actually using
> configurations like that we could choose arm1176jzf-s (Raspberry Pi
> has this, so it's reasonably common at the moment). In my opinion we
> can forget about arm9 and earlier for this, so unless anyone pipes up
> to disagree I'd be happy to commit the arm1176jzf-s one.
> 
> As for HasHardwareFloat, I think I'd just go for
> "Triple.getEnvironment() == llvm::Triple::GNUEABIHF", I don't think
> any of the others need hardware float.

Ok, thanks for the guidance. I have implemented that in my patch.

> It's probably best to post patches as an attachment to cfe-commits
> rather than cfe-dev too (more people read that list for reviews and so
> on). 

I'm afraid that does not match my experience.

But as I have your attention, I guess you'll commit it eventually, when the 
tests are in place. I can post it to the other mailing list too, but as far 
as I see, people active in (at least post-commit) reviews there are also on 
this mailing list.

> Tests are also a good idea. See
> http://llvm.org/docs/DeveloperPolicy.html#making-a-patch

I have no idea how to unit test this change. Any guidance on that?

Thanks,

Steve.