On Thu, Jun 01, 2023 at 01:26:13PM +0200, Laszlo Ersek wrote:
On 6/1/23 13:18, Richard W.M. Jones wrote:
[Sorry, not read the other replies yet, in meetings all day today]
On Wed, May 31, 2023 at 05:13:58PM +0100, Richard W.M. Jones wrote:
To start with I have opened a discussion topic. We can think about opening bugs later.
https://discuss.ocaml.org/t/ocaml-heap-fsck-and-forcing-collection-of-unreac...
There were some replies to this thread. One open merge request is notable - it adds compaction back:
https://github.com/ocaml/ocaml/pull/12193
although only for "small" blocks, less than 128 * 8-byte words, which probably covers everything we do.
We likely don't have a structure type larger than 1KB :)
It seems like full_major is the way to force freeing of unreachable objects, in all versions of OCaml.
Ah, does that mean we don't need a version check here, after all? IIRC, you said that in OCaml 4, "compact" was a superset of "full_major". If "full_major" is sufficient for finalizing unreachable objects on OCaml 4 too, then we can downgrade "compact" to "full_major" regardless of OCaml version.
Yes .. we use Gc.compact in at least two different ways. Only where we want the "free unreachable" meaning, replacing it with full_major does seem sensible. But we have to be careful not to replace the other one ("do an fsck") at least until we understand the best way to do that. Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com libguestfs lets you edit virtual machines. Supports shell scripting, bindings from many languages. http://libguestfs.org