[aarch64-port-dev ] macOS/Aarch64 workflow

[Andrew Haley]

On 01/10/2020 16:00, Anton Kozlov wrote:
>
> On 01.10.2020 11:51, Andrew Haley wrote:
>> It is more conventional, and IMO it would make more sense, to use a branch
>> in the aarch64-port repo. Ops at openjdk is working on it.
>
> I have looked through my mailbox, I hope you mean an aarch64 git
> repo (and not the hg one[1]).

I do.

> But I'm not sure this was discussed before in public.
>
> I see no benefit of an aarch64 branch, now you have to be an aarch64 project
> committer to push there.

Oh, is that a problem? That seems like no big deal to me, but I
suppose it is a couple of weeks' wait.

> More, there are downsides. In jdk-sandbox, you can have multiple
> branches for the same purpose. I'm aware of at least three
> implementations that do share a lot of in common. It will be very
> useful to see them side-by-side and pick good and bad parts in
> each. Second, it will be possible to commit something wrong and then
> fix the history (if we screwed up and it prevents reviewing -- the
> point of having a pile of fixes in public). We can also relax a bit,
> and avoid creating a Jira bug for each commit.

We've never used bug IDs in the aarch64-port project.

I guess I can see the appeal of a "Wild West" approach to this, but it
is a new way of working.

One of the advantages of the aarch64-port Hg forks is that we've got
the entire history of the project, back to the first code we ever
executed:

http://hg.openjdk.java.net/aarch64-port/jdk7u/hotspot/rev/9a723c72df9e

and the day four months later we got "Hello, World!" working, some
750k bytecodes:

http://hg.openjdk.java.net/aarch64-port/jdk7u/hotspot/rev/442027967d0a

This history is frequently useful.

Of course, you can work any way you want; it's entirely up to you. As
long as the patches you produce are clean enogh to review and not so
huge that they are too much to digest, that's all you have to do.

> Sometime later, we may need a repo carved in stone, but at such an
> early phase, I don't see a real value here.

Maybe. I guess the history in the sandbox will be easy to find in the
future, but who knows? I do know that anyone coming to
http://hg.openjdk.java.net/aarch64-port can easily find the entire
history of the port.

-- 
Andrew Haley  (he/him)
Java Platform Lead Engineer
Red Hat UK Ltd. <https://www.redhat.com>
https://keybase.io/andrewhaley
EAC8 43EB D3EF DB98 CC77 2FAD A5CD 6035 332F A671