Re: [Gnash-dev] Test failures in Date.as

[John Gilmore]

> These are dependent on floating point optimizations. Generally the
> failures are due to whether an intermediate value in the calculation are
> stored or not, and that depends on GCC version and CPU type. It's not
> something that C++ defines a semantic for. There are some compiler flags
> to fix it, but in the end it may be more sensible to drop the tests and
> accept that very large numbers may produce different results.

If ActionScript defines the result to be a particular value, then the
gnash implementation is buggy and these should be failures.  If it
doesn't, but the tests are too picky, make tests that don't depend on
the edge cases of Flash's float or double implementation.

        John