[Raspbian-devel] Confusion and decay around the serial UART - need help

[Eric S. Raymond]

Diederik de Haas <didi.debian at cknow.org>:
> On dinsdag 25 oktober 2016 14:11:33 CEST Eric S. Raymond wrote:
> > Rather than putting up that warning, which would have the inevitable
> > side effect of making Raspbian look like an unstable toy, I'm asking
> > for fast help repairing the recipe.
> 
> Have you tried to add enable_uart=1 to config.txt?

Yes. Done alone, it didn't fix the problem.  Supposing it had, I would still have
had two other issues with broken PPS and my udev rule no longer working.

I'm actually quite prepared to believe that enable_uart=1 might be
part of a fix, but apparently some other magic needs to be done as
well and I don't know what that is because *nothing is documented*.

I found enable_uart=0 in my /boot/config.txt after asking raspi-config to
disable console getty.  But this doesn't seem right - not only was agetty still
running on /dev/tty1, but I could find no way to stop it.  The obvious systemd
incantations didn't work.  I eventually decided to ignore *that* problem,
since I'm really only interested in /dev/ttyAMA0. But it points at yet another
bug or mis-documentation - raspi-config says it's going to disable login
on serial, then instead of actually doing that disables the UART and
leaves agetty still running.  Is /dev/tty1 the console now?  Is AMA0?

Who can tell, because *nothing is documented!*.

After sending my previous email, I tried putting a known good 2016-03-18
image on SD, then skipping the apt-get update step so it wouldn't download
a newer kernel that broke UART/PPS.  Failed, because someone decommissioned
the repositories that supported that image.

So now I and the HOWTO's would-be users are truly screwed. There's no
fallback path that makes an image that works as well as 2016-03-18 did
when it shipped.

This is a major regression and *seriously* disappointing.  I have put up
the following warning:

   *Warning: due to severe breakage in Raspbian, this recipe no longer
   works.  The maintainers apparently lack either ability or will to
   maintain stable, documented interfaces to the serial UART and 1PPS
   facilities it needs. Attempt only if you think you have the technical
   skill to figure out what they have busted and fix it.*

   *And no, you can't get around the problems by starting from a known-good
   2016-03-18 image. They took down the repositories you need for
   NTPsec's build dependencies.*

If that seems harsh, why has so basic a thing as a trivial udev rule
stopped working?

Where is the documentation on what "the console" is?  Is it /dev/ttyAMA0?
Is it /dev/tty1? Are these device somehow identified now?

Are there new, known bugs in pps-gpio that would explain why it stopped
working?  If so, is there any plan to fix them?
-- 
		<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 811 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/raspbian-devel/attachments/20161025/6c794736/attachment.sig>