• Iain Buclaw via Digitalmars-d (11/18) Sep 27 2014 Moving onto phobos std.math falls into this trap too, eg:

Iain Buclaw via Digitalmars-d <digitalmars-d puremagic.com> writes:

On 26 September 2014 01:32, Martin Nowak via Digitalmars-d
<digitalmars-d puremagic.com> wrote:
 On 09/24/2014 11:35 PM, Iain Buclaw via Digitalmars-d wrote:
 Bonus points, D inline assembly in a nothrow function.


 Marking asm as throwing by default doesn't make too much sense IMO,
 because it's fairly difficult to throw from iasm (never seen it) and the
 main point about iasm is full control.
 So why not simply trust the programmer decision here?

Moving onto phobos std.math falls into this trap too, eg:
getControlState/setControlState being culprits that affect me (I wrote
non-asm paths for pretty much every other std.math function) should I
choose to continue to disallow gdc asm in nothrow functions.

Though it is quite clear that all asm statements in std.math are
nothrow, what would be nice is a  trusted or nothrow attribute to be
added to asm statements to denote validated inline assembly in nothrow
functions.

Iain.