Closed facundominguez closed 7 years ago
@facundominguez Please resubmit the PR into the LeapYear/sparkle repo so that we can track it with ZenHub, assign it, and merge it as we desire.
@facundominguez Please also assign the PR to @alpmestan as a reviewer.
(Also, I think that if you entitle the PR "Fix #85", then when the PR is merged this issue will automatically be closed.)
On @mboes behalf:
Fixing finalizers for global references in jvm is looking in good shape. I pushed a few changes to @facundo's branch that make the patch a bit smaller. Mostly by exploiting a new feature I contributed to inline-c today. All unit tests pass, but we need to find a good solution to @facundo's last comment in #28. Namely: we need to make finalizers a noop once the JVM itself has been finalized. I have actually observed in practice. It's relatively benign: it just means you'll sometimes get segfaults when everything is finished and you try to terminate your program, but it shouldn't affect computations. Also, I'll spend some time tomorrow to document the object lifecycle invariants I think we have converged to.
The PR in tweag/inline-java#28 was merged. Is there more work to do in this issue?
It remains updating sparkle. We did it on our fork but not upstream. This will require either making a submodule of inline-java, or waiting for @mboes to release new versions of jni
and jvm
.
@mboes What do you think?
This has been addressed upstream already.
Requires https://github.com/tweag/sparkle/pull/97, actually.
PR for inline-java https://github.com/tweag/inline-java/pull/28