On 2012/04/17 14:26:23, Yang wrote:
On 2012/04/17 14:17:48, Net147 wrote:
> On 2012/04/17 14:02:51, Yang wrote:
> > On 2012/04/17 13:54:26, Net147 wrote:
> > > On 2012/04/17 13:44:56, Yang wrote:
> > > > On 2012/04/09 07:36:46, Net147 wrote:
> > > >
> > > > Please explain why this is necessary. Is the implementation of
pow in
> > > MinGW-w64
> > > > somehow faulty? It seems to me that you are just handling a few
special
> > cases
> > > > differently. Are you sure you caught all special cases where pow
would
> > return
> > > > the wrong result?
> > >
> > > MinGW-w64 has a custom implementation of pow that behaves slightly
> differently
> > > to MinGW, MSVC and GCC on Linux in these special cases. I've
handled all
the
> > > cases that are tested in the V8 unit tests. With this change and
previous
> > > changes, all unit tests pass with MinGW/MinGW-w64 4.5.2 including
> > > https://code.google.com/p/v8/issues/detail?id=1063.
> >
> > I see. I think it's more important to make pow compliant with the ECMA
spec,
> so
> > it might be better to test against test262 as well. Also, there is one
more
> > place where pow is used (Runtime_Math_pow in runtime.cc), which
should be
> fixed
> > as well?
>
> Runtime_Math_pow just calls power_double_double in assembler.cc which is
handled
> already. How do I run test262?
My bad. You can find instructions for test262 in test/test262/README.
All 34752 unit tests from test262 pass. Quite a lot of tests. I ran the
tests
using 8 shards/threads to speed it up testing.
https://chromiumcodereview.appspot.com/10026017/
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
