From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <219d633c491c593eafe8c551f142751d@hamnavoe.com>
To: 9fans@9fans.net
From: Richard Miller <9fans@hamnavoe.com>
Date: Fri, 20 Sep 2019 09:43:00 +0100
In-Reply-To: <20190919224632.33A4A1570CEA@mail.bitblocks.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Subject: Re: [9fans] go under plan9 on the radpberry pi?
Topicbox-Message-UUID: 091b7024-eada-11e9-9d60-3106f5b1d025

> Only lightly tested.

In a sense, plan9/arm go is tested as well as any other platform:
under the go continuous development process, every time a change
is made to the compiler or runtime library, a complete test suite
is run on builder machines for every supported architecture and
operating system.  If you look at https://build.golang.org and
scroll wayyyyyyyy over to the right, the plan9/arm column refers
to a set of Raspberry Pi machines run by David du Columbier and me.

In another sense, it's probably not very well tested at all:
I'm not aware of any production application being run on go in
Plan 9, on any machine architecture.  I haven't used go seriously
myself, but I find the test suite gives the OS such a brutal workout
(especially with small physical memory) that it's a good way
to flush out underlying Plan 9 bugs.

The tests show some intermittent hard-to-reproduce failures ("flakes")
on all the Plan 9 builders.  Many are timing issues because the tests
make assumptions about absolute speed of builder machines; but there
are some "can't happen" panics during garbage collection which smell
like a cache or memory barrier problem.  Please don't use plan9/arm
go to run your nuclear power plant just yet ...