[fpc-devel] Subject: six issues from a native aarch64-win64 port (three of them not target specific)

[Sven Harazim]

No rush at all - and sorry about the SD card.

If it saves you some of that: the hardware is set up here, so I am happy to run
anything you want checked on aarch64-win64 and report back, patched or
unpatched. Two full trees are already built and installed side by side, so
with/without comparisons cost me a make run rather than a setup.

Two of the six do not need ARM hardware at all, in case you want to start where
your test system is not in the way: 41862 (optcse, needs only -OoNOREGVAR or an
inlined label) and 41865 (the Currency regression, any target).

One thing I added to 41870 after sending the mail: it is a regression, from
b60874ba67, in trunk since 6 July as part of !1268. That commit removed a
comment saying the epilog scope index has to be 0 and then set it to 1, so the
patch only restores what the comment described. No release is affected.

And if it is ever useful: aarch64-win64 has no automated coverage at all today,
the CI being Linux only. I can run the testsuite here regularly and send the
digests, if that path is still open - that costs me nothing and does not depend
on my machine being up at any particular moment. A gitlab-runner would work too
in principle, there is a native windows-arm64 binary, but this is a laptop, so
it would only be honest as a scheduled job rather than something MR pipelines
wait on.

Regards,
Sven Harazim