Jan Midtgaard

Results 395 comments of Jan Midtgaard

I've now created a reproducer and filed this upstream: https://github.com/ocaml/ocaml/issues/13586

This just triggered again on 32-bit `5.3.0+trunk` by the merge to `main` of #460: https://github.com/ocaml-multicore/multicoretests/actions/runs/9169655398/job/25210472949 ``` random seed: 103830913 generated error fail pass / total time test name [ ]...

First off, this is a collection of failures we observe. Once we have fleshed out reproducible steps, these are reported upstream. Help is very welcome, snarky remarks less so. >...

OK, fair enough. Some of these remaining ones are just hard to reproduce - I suspect because they are timing or signal related. I've been trying today for this one:...

I finally managed to reproduce this one - on 5.2.0 - and only once for now. It is indeed a sequential fault! :open_mouth: https://github.com/ocaml-multicore/multicoretests/actions/runs/9180414541/job/25244781126#step:18:762 ``` Starting 74-th run random seed:...

Switching hard-coded seed to 107236932 (the first one) works much better! Across 500 repetitions this triggered 56 segfaults on `5.2.0` https://github.com/ocaml-multicore/multicoretests/actions/runs/9181424522/job/25248130724 and 50 segfaults on `5.3.0+trunk` https://github.com/ocaml-multicore/multicoretests/actions/runs/9181424534/job/25248130821

I've made a bit of progress on this. - The issue still triggers on 5.2.0 and a fresh `trunk` as of this morning on the CI - I'm still unable...

I've tried testing previous versions (5.0.0, 5.1.0, 5.1.1) too - both with a hard-coded seed 107236932 and with random ones. Result: - 5.0.0 - not able to trigger - 5.1.0...

> How easy/hard would it be for you to run the testsuite on an arbitrary patch, if I send you changes to the runtime that might be related to the...

No cigar unfortunately: https://github.com/ocaml-multicore/multicoretests/tree/domain-dls-32-bit-gabriel https://github.com/ocaml-multicore/multicoretests/actions/runs/9352909686/job/25742143625 On that one I also saw this (even rarer) non-crashing misbehaviour: ``` Starting 89-th run random seed: 135228812 generated error fail pass / total time...