Stas Sergeev

OK, I know what to do! :) Lets not drop the tests for now. Instead, we need to add a few new failure categories. For example: "unstable-on-msdos" and "unstable-on-fdpp". What...

> but the CI should > be considered "passed", unless some > fails on both DOSes. I realize this is a bit of cheating, but: - its still better than...

Also if we eventually notice that all the unstable failures are fdpp-related and no failures observed on ms-dos, we will be able to tweak the technique and keep marking any...

A further improvement may be to add an "msdos+comcom64" test suit. It may not run by default, but run only in case some failures were seen on fdpp+comcom64, which have...

But I really think that without such an advanced probes, there is no chance to find the instability we currently obviously have.

Looking at this log: https://github.com/dosemu2/dosemu2/actions/runs/9085777250/job/24969889885 I think each libi86 test runs lots of config probes for dosemu2, each one running dosemu2 itself: ``` | if "$DOSEMU" $DOSEMU_OPTS -tde \ |...

Yeah, would be good to duplicate 5-10 tests, starting from those that recently failed. Also we can have 1 "dummy" test, i.e. the one where dosemu2 probes for lpt/com etc...

This branch seems to have conflicts. Andrew?

Thanks! I think off-topic came from some CI problems that are already solved.

This is probably a false-positive. Its just a sign-extension op: we assign from smaller field to a larger one of the same register. Please try to experiment with different syntaxes....