Jan Midtgaard

I've spent some time on this PR today: - I refactored the code into a `Model` module with relevant operations - I restructured the error cases in `postcond` to be...

Rebased on main as I wanted to get an idea of whether this helps on #359. I expect the Windows runs to still fail. Also (as a note to my...

Rebased yesterday to trigger these on the patched `5.1.0~beta1` release. There is clearly some way to go :grimacing: While deciphering errors, I've identified that this has a side-effect of removing...

I did a fresh rebase of this PR. As part of it, I removed the pattern-matching on specific error messages, as I found this was digressing into reverse-engineering of Linux/macOS/Windows-specific...

I pushed a clean-up commit d2a1345 which - accepts the Windows rename regression - quantifies all "accept workarounds" with OCaml versions CI summary - 32bit trunk found an unexpected counterexample...

Rebased on `main` with conflicts resolved

CI summary for f8dc3b9 - this is still running focused Sys STM tests: - 32bit 5.2 - 0/10 attempts triggered a parallel counterexample - 32bit trunk - 4/10 attempts triggered...

Thanks Olivier! Quick observation: why does this cause so many `Failure("failed to allocate domain")` errors? Overall, this should allocate less `Domain`s - not more. Are there conditions under which the...

Thanks again for this! :pray: - From your numbers it seems we should be able to get a decent speed-up, even if we just retain the pool to the property....

I think I figured this one out too: the hint was `STM` tests with non-trivial `precond` triggering `failed to allocate domain`. Preconditions are checked before Domains running, and if the...