Does MacPorts provides its Test/QA results online?

[Jeffrey Walton]

On Wed, Oct 28, 2015 at 10:19 PM, Brandon Allbery <allbery.b at gmail.com> wrote:
> On Wed, Oct 28, 2015 at 10:15 PM, Jeffrey Walton <noloader at gmail.com> wrote:
>>
>> The immediate problem here is GCC supplies an option, documents the
>> option, and then the compiler rejects it. Also, if its is an LD issue,
>> then I believe LD should reject it. The compiler should not reject it
>> by proxy.
>
>
> Then you need to convince Apple of that because they supply the linker. The
> current version shipped by Apple accepts and ignores the option. Versions of
> gcc that take advantage of this are subtly wrong as a result. Given that
> Apple does not bother to complain about options it does not implement, it's
> considered better to error out than to silently produce possible runtime
> bugs. Other providers of gcc apparently feel that subtle, silent runtime
> errors are acceptable.

Just to play devil's advocate: what about projects that use linker
scripts to ensure the order of the static constructors? They don't
suffer Apple's LD problems. Why is it good policy to deny them and
fail their build?

You also addressed the warning. Perhaps it would be better policy for
the compiler to issue a warning rather than an error. (I happen to
like this idea, but its just bike shedding on my part).

For what its worth, I take the problem very seriously because it
results in {occasional|rare} non-deterministic failures. I was going
to provide a linker script for Apple platforms. But we have to get
past the compiler problems first.

Jeff