[cfe-dev] Small patches to allow fully independent clang/llvm/compiler-rt/libc++

[C Bergström via cfe-dev]

On Thu, Oct 15, 2015 at 12:10 AM, Renato Golin <renato.golin at linaro.org> wrote:
> On 14 October 2015 at 17:59, C Bergström <cbergstrom at pathscale.com> wrote:
>> I don't know why gcc does it, but gcc -v will give you the configure
>> line which was used to build the compiler.
>
> GCC does it because the back-ends were never meant to work on the same
> code. That was a decision taken many years ago and the make flags is
> just a consequence of that.
>
> For the last 2 years Linaro was trying to make the GCC ARM and AArch64
> back-ends work together, but so far, haven't succeeded. I don't want
> to us to get into that situation.
>
> I know that's not what you're proposing, but Jon's concerns about
> testing and overall design are indeed compelling. And this opens up to
> changes that will get us in a place that is not exactly like GCC, nor
> current LLVM, which is probably worse. Moreover, the proposal of
> having an LLVM toolchain actually solves most of the problems and is
> indeed on par with the rest of the driver's design. I agree it's the
> best solution.
>

*some* default is already being decided. you must use a flag to
override that... the flag to overwrite that default can still exist.
(I don't think the patch removes that??) The driver doesn't decide
*everything*. These arguments aren't logical. Adding an optional
compiler time flag can be benign and certainly doesn't cause a
negative impact on test or QA inherently.