[llvmlinux] Added patch for detection of compilers that support unwind for ARM. Now clang is detected.

Behan Webster behanw at converseincode.com
Wed Jan 29 17:30:47 UTC 2014

On 01/29/14 08:20, David Woodhouse wrote:
> On Tue, 2014-01-28 at 10:34 -0800, Behan Webster wrote:
>>> You couldn't send that tree to Linus today, or even put it in linux-next
>>> without triggering a lot of complaints, surely?
>> Indeed. Which is why we haven't. We don't have proper solutions for
>> everything yet (though we're getting closer). The project freely admits
>> that there are still some work-arounds still looking for a better solution.
> And that's absolutely sane. But I think it's useful to keep tabs on what
> we do have — where we have patches which are sane for upstream they
> should be pushed that way as rapidly as possible, and where we have
> workarounds for LLVM they should clearly identify the PR# that they
> address.
Understood. I will endeavour to fix this.

>  Looking through the patches we have at the moment, it's
> sometimes hard to tell what's being fixed, and why.
Understood. I'm not sure we always update comments when we "fix" our
quilt patches. I'm actually just going through and fixing descriptions
right now.  I will add LLVM/PR numbers where possible.

> For example,
> /arch/i586/patches/9004_fix_for_edx_register_on_64_bit_as_suggested_by_hpa.patch
> gives me *no* context about what the problem is or where to find the
> existing discussion about it.

>>> There's definitely merit in sending the simple and straightforward
>>> *fixes* upstream under separate cover, as quickly as they're generated.
>> Then we should start doing so. Thanks for the suggestion. Can I ask you
>> to be a reviewer for such a repo?
> Yeah, absolutely. If you want to let me have simple patches with a clear
> commit comment I can certainly shepherd those upstream.
Awesome. Thanks.


Behan Webster
behanw at converseincode.com

More information about the LLVMLinux mailing list