Hi Robert,
On 9/24/21 6:02 PM, Robert Wilhelm wrote:
Hi Jacek,
thank you for the review and the testcase.
It seems it cannot be solved in interpreter alone, because bytecode for y=f and y=f() is same in following example.
I would be glad on advice how to fix this...
Something like the attached patch should take care of required compiler changes. The interpreter part just replicates the existing behaviour, but it should be possible to implement proper behaviour on top of that.
Thanks,
Jacek