Re: Anyone using stretch/buster/sid on ARMv4t ?
- Date: Tue, 7 Nov 2017 13:43:50 +0100
- From: Héctor Orón Martínez <hector.oron@xxxxxxxxx>
- Subject: Re: Anyone using stretch/buster/sid on ARMv4t ?
>>>[+debian-embedded, feel free to adjust CC'd mailing lists on reply]
Thanks for bringing up this discussion! And apologies for adding
more complexity to the initial question.
Find few comments inlined below,
2017-11-05 22:32 GMT+01:00 Adrian Bunk <bunk@xxxxxxxxxx>:
> for the armel port in buster the question of raising the baseline came up.
That has been a recurring question over the time, the reason to
maintain ARMv4t instruction set was OpenMoko mobile phone, which lot
of people was using back in the days.
I am unsure that is still relevant. However there are industrial
products based on old ARM CPU, but they are probably good to bump to
> Roger Shimizu is doing a great job on ARMv5 hardware and I've seen bug
> reports from users on ARMv5 hardware in stretch, so it is clear that
> ARMv5 should stay supported (and of course also ARMv6 and ARMv7).
There are important issues to keep the port going forward, and it has
been flagged as deprecated in last couple DebConf's ARM BoF meetings:
Also build daemons are aging, those were initially donated by Marvell
(some development boards) which then they replaced with other
development boards. We have been unable to find suitable hardware to
build armel port and current ARMv8 server class instruction set lacks
few instructions and we have been advised to not build armel ports in
Current build daemons got into production in 2010 --
We have no current replacement for that hardware, which it's already 7
years old and still needs to keep up running for Stretch life cycle
(and maybe LTS).
There might be other issues, as upstream support in few other projects, etc...
(For build daemon issue discussion, please feel free to fork the
thread and discuss at debian-arm mailing list.)
Having outlined few part of current issues, we had a proposal in dc16:
>>> * Partial armel architecture?
No clear plan for how to create and maintain a partial
architecture, let alone releasing.
Drop desktop tasks and dependencies.
Keep web server and mail server - trivial in comparison.
Lack of video use cases.
Headless partial arch for armel?
Build daemons would be a problem, so let armhf buildds do
the builds. Can't do that on ARMv8 (which can do armhf).
Kernel helper to emulate barriers but hardware support can
be lacking on ARMv8 hardware.
>>> * Update from v4t to v5tel and go headless?
No kernel support for exclusive v4t.
Openmoko would be ruled out of this partial arch.
Kirkwood and Orion can be maintained.
openssl and other updates would be useful within stretch.
Propose to keep in stretch and go to v5tel partial arch after stretch.
Not clear how much work it is.
Packages would have to specify armel or it will not build.
In other words, as I see it, there are few tasks we could work on
(with added problem of lack of resources):
A) Select a package set to provide headless support, not only for
armel, but also for powerpc and other non-existent Debian
B) Setup cross build daemon and cross build such set of packages,
provide build logs and fixes to package maintainers.
C) Follow up on bootstrapping work and make sure we are able to sanely
bootstrap architectures from cross built packages or bootstrap it from
D) Talk to release team and other involved teams and design a way to
support (cross built) partial architectures in official Debian.
Note that the above can be huge amount of work and it might need
skilled people as well as funding to become a reality.
In a way that's my personal vision to keep aged architectures around
supported in Debian and dropping our hardware dependency, and getting
also nice features that other parties might find interesting such
cross building support and bootstrapping.
(For technical discussion on cross building, please fork the thread
and CC debian-embedded.)
Héctor Orón -.. . -... .. .- -. -.. . ...- . .-.. --- .--. . .-.