On Mon, Jul 18, 2011 at 5:52 PM, Diego Novillo <dnovi...@google.com> wrote:
> On Mon, Jul 18, 2011 at 20:45, Gabriel Charette <gch...@google.com> wrote:
>> To demonstrate my point:
>> add /* { dg-options "-ffinite-math-only -fno-math-errno" } */
>> to c1builin-integral.cc
>>
>> and watch the test pass (since now we are using the flags across all
>> compilations).
>
> Ah, OK. That's your fix then.
>
Ya, that fixes the test.
But the sole fact that pph was not throwing an error is wrong in my opinion.
We should probably ensure that the same flags are enabled/disabled
that were at the pph compile time (otherwise we could do the parsing
with one set of flags and compile with a different set of flags...)
We probably only want to enforce this for a subset of the flags (i.e.
we don't care about flags like -Wall and stuff like that, but only
about the flags actually affecting the binary output).
We can discuss this in the Hydrazine meeting tomorrow.
Gab
