r230160 - Relax the requirement on sized deallocation a bit: Default on unsized delete if sized delete is not provided in global scope, and -fdefine-sized-deallocation option is disabled.

[Renato Golin]

On 23 February 2015 at 11:07, Timur Iskhodzhanov <timurrrr at google.com> wrote:
> Green bots are no doubt better than red bots or no bots, but I'm not sure
> what level of ARM support our team has committed to.

I appreciate that, but you're not the only sanitizer developers out
there. If we found a bug in ASAN for ARM, we have to fix it. If that
means involving other groups / companies (like Linaro), than so be it.

You may not have committed to any ARM sanitizer work, but we have. So,
if there is an issue with one of your patches, we must help you solve
them. And, in the same way as any other project in the LLVM sphere,
you can't just XFAIL tests on another platform just because you
haven't committed to working on that platform.

Two years ago I have fixed all test-suite failures for ARM, and
coincidentally, I have also fixed it for PowerPC and Intel. On one
occasion, I have reverted my change and worked with other developers
to make it work on those platforms, too. You can imagine that I also
haven't committed to work on Intel or PPC, but that's what it takes to
get the product stable. Overall, I don't trust in my solution if it
only works for my own architecture. In the long run, it's better for
you if it continues working for ARM.

However, despite all of that, it's our community's policy to not hide
failures under the rug, especially if it concerns a first-class
architecture, like ARM. The sanitizers are not a local project any
more, they're part of our community, and we have to care for it like
we care for all others.

cheers,
--renato