On Fri, Jun 2, 2017 at 9:09 PM, Marcus Denker <marcus.denker@inria.fr> wrote:
The first thing we need is an automatic test run for 64 bit���


has just 32bit (win, linux, mac). It needs to have 64bit, too.

But I think we should not wait with the release for that.

In that case maybe we should call 64-bit a technology preview, and push for an early point release that fully passes tests - maybe worth a sprint focus.
cheers -ben��
��

On 2 Jun 2017, at 14:27, Sven Van Caekenberghe <sven@stfx.eu> wrote:

Hi,

I am wondering a bit about the state of Pharo 6 64-bit.

On one hand the current RC seems to work really well on macOS (for me), it is stable and clean.

Is the goal to ship this variant as a real finished version ?

If so, I think must make it pass as much unit tests as possible.


I filed a couple of issues related to that:

https://pharo.fogbugz.com/f/cases/20102/SmallInteger-digitAt-not-ready-for-64-bit-Some-Kernel-Tests-Numbers-fail-to-take-64-bit-into-account

https://pharo.fogbugz.com/f/cases/20108/CompiledMethod-isAbstract-fails-on-64-bit

https://pharo.fogbugz.com/f/cases/20111/Some-Collection-tests-fail-on-64-bit-because-they-assume-Floats-are-non-immediate

https://pharo.fogbugz.com/f/cases/20110/AllocationTest-testOutOfMemorySignal-not-well-suited-to-64-bit

The first one has a fix. Fixes must be done carefully, since the same image must be able to run on both 32 and 64 bit.


I also noticed that a couple of Compiler tests are failing. Now, obviously the compiler works good enough since it can load code, so this is probably also something minor/weird/unexpected. About 51 tests fail (in MethodMapTests, OCBytecodeHGeneratorTest and OCNewCompilerWithChangesFunctionalTests). Should I create an issue ?

Sven