Regis wrote:
Tim Ellison wrote:
Regis wrote:
Tim Ellison wrote:
Regis wrote:
RI has the different behaviors on Linux and Windows in this test case,
and Harmony has the exactly the same behaviors, so I think our
implementation is OK.
And there are some other similar cases in SocketTest, which failed on
Linux but passed on Windows, I think it's why it is in
exclude.linux.x86.drl. If we fixed them, SocketTest could be moved
from
exclude list.
It wasn't the fact that there is a difference that I object to here,
but
the way you have achieved it does not fit in the Harmony architecture.
These types of differences are pushed into the native code. I believe
that this should be done here too.
Our tests are all written in java, I think it's better to keep it. How
about move these platform depended tests to platform directory, like
UnixSocketTest or WinSocketTest?
D'oh, sorry, I had missed that they are tests!
The platform specific tests are sorted into directories under
src/test/api (Java API-based tests)
or
src/test/impl (Harmony impl types tests)
The test you modified is in
src/test/api/common
but if it is platform specific, the relevant parts of the test should be
in one of
src/test/api/windows
or
src/test/api/unix
Does that make sense?
Regards,
Tim
All right, I'll move them to platform folder.
The new fix committed at r766631. I'll continue tide-up SocketTest, and try to
move it out from exclude list.
--
Best Regards,
Regis.
