gcc can target wrong instruction set when building JDK native code

[Martin Buchholz]

On Mon, Apr 7, 2014 at 10:48 PM, David Holmes <david.holmes at oracle.com>wrote:

>
> So gcc's behaviour seems at odds with its documentation. But regardless we
> should not be relying on whatever default is used but should set
> -mtune/-march explicitly as is done by hotspot.


I agree. At the start of every release, the project should decide how far
back to support old hardware and define default mtune/march flags for
supported platforms.  For unknown platforms, take the default of the
compiler.  But it's a fair amount of maintenance work, since these flags
are both compiler and platform-dependent.

I'd be curious to know what the project policy is on supporting old
hardware and old OSes.