Recent RPi dev builds won't boot

I used to update my RPi 3’s version of RISC OS fairly often, but my last try (in August) and today’s try won’t boot:

Machine startup has not completed successfully: ‘Route: add net default: gateway 192.168.16.1: Network is unreachable’

I updated Start.elf, Fixup.dat and Bootcode.bin before I did today’s upgrade, and checked that the machne still booted with the existing version of RISC OS before I updated Riscos.img.

What have I done wrong, or not done? Any help is welcome.

This machine is also the one that started falling over when the new IPv6 stack went on it, so none of the changes for the experimental stack is in there, if that makes any difference.

Has the IPv6 stack started going into dev builds? My attention has been elsewhere.

When I run !Boot.Choices.Internet.Startup I get:

ifconfig: RasPi3B: bad value

RasPi3B is the machine’s name, and is in the Hosts file with the value 192.168.16.69

Surely the local host name, if in the hosts file, should be listed against 127.0.0.1
Although I’m pretty sure it doesn’t need to be in the hosts file anyway, like most of the entries anyone ever puts there :)

Machine startup has not completed successfully: ‘Route: add net default: gateway 192.168.16.1: Network is unreachable’

Don’t worry, this is a well known error and arises when you set the gateway address but are using DHCP. Try unsetting the gateway address or using static IP addresses throughout. One day this bug will be sorted…

This sounds more like you still have bits of the new Network stack installed. This you use the offical uninstaller to remove it?

I solved the problem by moving from obtaining IP address from hostname to manually with an explicit IP address.

I take it that “from hostname” simply doesn’t work with the new IP stack?

I solved the problem by moving from obtaining IP address from hostname to manually with an explicit IP address

I create reservations¹, so the DHCP allocated IP is the same every time.

¹ I do that at work too, sorry wrong phrase – I long ago set the rules so that everyone at work sets the machines as DHCP and either personally, or asks someone with access to set reservations.
Most equipment manufacturers tend to argue of course.

My computers and some local gizmos tend to use static IP. The printers, tablets, and phones use DHCP. The router maintains a list of DHCP allocations both for the DHCP devices (so they get the same address each time) but also for the static IP devices (so these addresses won’t be handled out).

How does the hostname thing work? I get hostnames as outgoing references (google.com, heyrick.eu, etc), but for setting up networking in the first place?

How does the hostname thing work? I get hostnames as outgoing references (google.com, heyrick.eu, etc), but for setting up networking in the first place?

I suspect it’s a holdover from the bad old (pre-DNS) days when every machine had to have a fully-populated “hostnames” file and some unfortunate person had the job of keeping this list up-to-date.

OK, it’s not all such good news from me after all. The machine starts up and I gan get mail and news, and I can use ShareFS shares. But LanMan98 won’t connect to either of my shares on the LAN (OpenMediaVault and WD MyCloud). The error is “Unable to resolve address: [hostname]”. This is bizarre to me as I can ping them by exactly the same names, from the command line.

I can connect if I give the IP address instead of the hostname.

Can anyone offer any insight?

(I disabled access to the MyCloud from the Internet, as advised by WD because they no longer provide security updates. Just in case anyone’s wondering.)

I take it that “from hostname” simply doesn’t work with the new IP stack?

That’s going to break the way I work if true.

I run a network of pi clones, offsite with no router present. The machines are all set up using “from hostname”. When I make a new configuration, I set one up, giving it a hostname outside the usual range, and save it’s card image. Then I produce copies of the card.

When the card it used in each new machine, I change the hostname. That’s it. The IP address is set from the hosts file, and the share name set from a little piece of basic which runs at startup.

To do that with DHCP would mean setting up a DHCP server on something, to replace the function currently on the router. What I psarticularly want to avoid, is using the router when at home, and something else when in “the field”. That’s asking for problems.

Where’s the hostname stored? A local file? If that’s the case, why not just use a static IP address?

@Alan: When your system is working, consider carefully any reason to upgrade ;-)

When your system is working, consider carefully any reason to upgrade ;-)

That’s a conservative viewpoint. I am firmly of the view that we have to keep things moving forward.

Where’s the hostname stored? A local file? If that’s the case, why not just use a static IP address?

My normal configuration for the network is to “Obtain IP address” “from hostname” (top left radio button). The hostname is in the Hosts file. I have had it set up like that for many years. Something has gone wrong during this year’s builds that makes such a setting not work.

In hindsight, I wish I had attempted to upgrade more frequently, and that I had followed it up when things started to go wrong.

The last three versions of RISCOS/IMG I have are 2022-05-08, 2022-08-21 and 2022-10-04. The first works as it always did; the last two show the symptoms I’ve described: obtain IP address from hostname doesn’t work, and LanMan98 claims to be unable to resolve share names to address.

My first guess is that the Resolver has failed in some way. I’ve got the 2022-08-21 version running at the moment, and “help resolver” shows the version to be 0.73 (16 May 2022). Interesting that it’s a few days after the date of the last RISCOS/IMG I have that works.

That’s a conservative viewpoint

I was really referring to Alan’s system that’s taken a lot of effort to get where it is.

There does seem to be an oddity for me with the ROD network stack (7.01) on my 5.29 ARMX6. Using DHCP (I use reservations for my main systems) all seems OK. With hostname in INetDbase:hosts, set ‘Obtain IP address’ to ‘hostname’, OK, but it’s stopped using DHCP completely (why???) so doesn’t pick up the route to the outside world. Then manually set default gateway and that’s game over for networking. [Edit: So I go to try to reproduce this issue just now to submit an error report, and it just works fine. Will leave in this configuration in case the issue resurfaces at some point.]

Stuart, could you report any oddities like that do ROD either via email or via the ticket system site please. I’m not sure how active the ticket site is now, but certainly if you can produce a report of the problem, we have active work going on the stack, so issues should be fixable.