Closed burner closed 5 years ago
don't remember the specifics, but I think removing malloc allocation from references.d is going to have some implications. Let's see here, for every D object that is wrapped and used in python, a reference to it is stored in Container, as is a reference to the python object. this prevents D from garbage collecting something python still has a reference to, and possibly vice versa. I don't think Container is malloc allocated to avoid gc references to objects exposed to D, that doesn't make sense to me. looking at https://github.com/ariovistus/pyd/blob/master/infrastructure/pyd/class_wrap.d#L95, it might be some similar fiddly implementation issue.
how do you suggest we get pyd compiling with 2.087 then?
well, if it was an implementation issue, it was 6 years ago. might not be an issue any more. guess we can try it and see what happens.
sounds like a plan, thanks.
Could we also get a new release
ya, we're a bit overdue for one of those
0.11.0 should be up now, let me know if it isn't somewhere
looks good so far, I'll let you know if there is something else.
Thanks
@ariovistus I think your fears were justified, I am seeing strange segfaults since this change. See https://github.com/ariovistus/pyd/pull/119 for an alternative
So the current way of working with allocators does no longer work. This PR pulls out the placement new's and delete's and the allocators.