Raspberry Pi 4

Out of curiosity I downloaded the RISC OS Pi SD card image, utilised a new SD card and an 8Gb Pi4 to test the new image.

Similarly afflicted by curiosity I downloaded ROOL’s Pi OS5.28 image, utilising balenaEtcher on a Mac to write an old 2GB card of dubious lineage (Maplin), and that worked perfectly in the 8GB RPi4.

My Pi 4 build is based on RC16 2020-06-14 and imaged on to a brand new micro-SD card on Windows 10, then updating the RISC OS rom images as they become available. Currently 5.29 22 Oct 2020. I have also updated fixup4/dat stamped 23 Oct 2020, start4/elf stamped 23 Oct 2020, bootcode/bin 12 Jun 2020, cmdline/txt stamped 18 Feb 2020.

Harinezumi – Will have to investigate that.

Sometimes it errors before the DHCP or during/after it. The worst sounding one was ‘Broken Directory’! But after a successful start a Disknight check says the card is OK. Other errors are: File ‘G’ not found. – Please insert card ‘RiscOSpi’. – File ‘SBPr’ not found.

My USB-C PSU is a 3 amp Stontronics one and sold with the Pi 4 as compatible. At that time the official Raspberry Pi PSUs had sold out. Once the Pi 4 has started up and running OK, I can plug into the USB port a Samsung SSD drive and use without any problems.

Reporter will also log boot commands and errors – see http://www.avisoft.f9.co.uk – but it does not change boot processing whereas I believe Harinezumi can enable errors to be ignored so boot continues.

This is a virgin !Boot as part of RC16

Reporter will also log boot commands and errors – see http://www.avisoft.f9.co.uk – but it does not change boot processing whereas I believe Harinezumi can enable errors to be ignored so boot continues.
bq. Reporter will also log boot commands and errors

Which is the next step after the simple logging and more resilient boot from Harinezumi :)

If Bernard follows the sequence of adding Harineumi, checking the log after each boot and establishes whether or not there is a repeating pattern pointing to a specific problem then we move to a fix or more detailed logging.
At the detailed logging point of the sequence I’m sure you can suggest a first step, a second and a third to gather more and more detail.

As I keep saying, horses for courses.
I’d suggest leaving Harineumi in the boot when the Reporter level investigation occurs.

Thanks everyone – One more symptom – A switch on startup will sometimes work OK without any errors. I’ve just had one again! I hope to try Harinezumi out tomorrow (Mon).

Sometimes it errors before the DHCP or during/after it. The worst sounding one was ‘Broken Directory’! But after a successful start a Disknight check says the card is OK. Other errors are: File ‘G’ not found. – Please insert card ‘RiscOSpi’. – File ‘SBPr’ not found.

Have you ruled out a faulty or degrading SD card?

Harinezumi – Now tried this on my Pi4 with 4GB RAM – Not compatible as it causes a memory error at computer startup and fails to proceed and reach desktop because of that. Obviously not compatible with more than 2GB RAM.

Reporter – After I corrected the Run file open problem, I could switch Reporter on. On startup this then lists the following error 1874 times:

[WindowManager/] Exec
WimpError – from Message from SDFS
Error – &00000000
Message – Please insert card ‘RiscOSpi’

Reporter try 2 – Other errors such as ‘Broken directory’ and ‘File SBPr not found’ are terminal and the desktop is not reached to view Reporter information.

I will probably try the new RISC OS 5.28 (Stable) image on another new micro-SD card next when I get time.

Not compatible as it causes a memory error at computer startup and fails to proceed and reach desktop because of that. Obviously not compatible with more than 2GB RAM.

I don’t believe that Rick has a machine with that much RAM, so can you post something more specific than ‘a memory error’?

I was going to try it on my own 8 GB Pi 4 but decided to install it on my Pi 3 first and then transplant the card. Alas, it doesn’t want to run on the Pi 3 either! But since this is the Pi 4 thread, I’ll post that issue elsewhere :)

I will probably try the new RISC OS 5.28 (Stable) image on another new micro-SD card next when I get time.

Another option might be to use the SSD or a separate USB pen drive, ADFS formatted, to be the equivalent of a HardDisc4, only using the SDcard to start the RPi. ;-) You might find that it improves the Pi experience.

the desktop is not reached to view Reporter information.

For that sort of case Reporter has a ‘log’ facility, where information is immediately written to a disc file. This is not for normal use, as it slows things down a little, but nevertheless can be useful when the desktop becomes unusable. However, in this case it sounds like it may be a disc problem, and the log needs a working disc…
Please contact me direct if you want any help with Reporter.

Another option might be to use the SSD or a separate USB pen drive, ADFS formatted, to be the equivalent of a HardDisc4, only using the SDcard to start the RPi. ;-) You might find that it improves the Pi experience.

This is exactly the setup here, and the improvement is noticeable. The following is an extract from my ROMark5 testing (SDFS stats first) which speaks for itself:

EDIT: I’m using a Filecore E+ formatted 240GB SSD. !Boot* plus all the HD4 standard directories are copied onto it.

[*Except the Loader partition of course – DON’T copy this!]

Other errors such as ‘Broken directory’ and ‘File SBPr not found’ are terminal and the desktop is not reached to view Reporter information.

I used to have trouble with SD card corruption. Somewhere, circa 2013ish, I wrote an epic rant about RISC OS’ inability to fix its own filesystem after it screwing up.

You need to do four things:

• Use a different SD card. They’re not identical and some just plain don’t work with the Pi. There’s a list around online of reports of cards that do/don’t work.

• Ensure your power supply is up to the job. If in doubt, change it.

• Check the cable between the power supply and the Pi. Some can be rubbish and deliver rather less than the desired 5V. If in doubt, change it.

• If you did anything in the config to fiddle with clocking, undo it. Get a working system and then tune it.

I did all of the above on my older Pi (1B) and all the corruption nightmares simply stopped.

Another option might be to use the SSD or a separate USB pen drive, ADFS formatted, to be the equivalent of a HardDisc4, only using the SDcard to start the RPi. ;-) You might find that it improves the Pi experience.

Are there any instructions anywhere on how to do this please?

I don’t believe that Rick has a machine with that much RAM, so can you post something more specific than ‘a memory error’?

Will send complete error message when I get home later. I did make a note of it on paper!

Are there any instructions anywhere on how to do this please?

See Moving !Boot to another drive

Yesterday I installed RISC OS 5.28 on my Raspberry Pi 3 including the files that go into !Boot.Loader. Today I wanted to test the RPi 4 (2GB) for the first time and inserted the same sd card into it. I thought that this is everything you have to do. Unfortunately I do only get a black screen – nothing happens even after several minutes. HDMI is connected to HDMI0 (also quickly tried the other port). I quickly tried out noobs to make sure the RPi is ok – but it looked all good and normal. So any help would be nice!

@Michael
The Pi4 requires:

1. Two extra pi firmware files (fixup4.dat & start4.elf) to be added to the DOS partition (i.e. within the Boot.Loader file)
2. extra pi4 commands to the config.txt file (these may already be present – but best to make sure)

Details → https://www.riscosopen.org/wiki/documentation/show/Software%20information:%20Raspberry%20Pi:%20Firmware

Ah, thank you very much, there was indeed something missing in the config.txt file. I write this on the RPi 4 – so works now ;-)

Harinezumi – As promised above, here is the Harinezumi error on a 4GB Pi4 with RISC OS 5.29

Error: Trap while in trap handler:
Internal error: undefined instruction at &0001AD9C, pc = 0001726C : registers at 0001D7DC (&800E00)
*Where &0001AD9C at offset &00012D9C in application memory.

See Moving !Boot to another drive

Thanks Stuart Painting

Internal error: undefined instruction at &0001AD9C, pc = 0001726C : registers at 0001D7DC (&800E00)
*Where &0001AD9C at offset &00012D9C in application memory.

You’re running version 0.09, yes?

Can you do this before running Harinezumi:

*Set Debugger$DumpOptions -file annotated
*Set Debugger$AnnotatedFile $.__haricrash

What this will do is instruct the Debugger module to write some useful information to the nominated file ($.__haricrash).

Then, after running Harinezumi:

*ShowRegs
*MemoryI PC -20 +40

Please paste the results of the ShowRegs/MemoryI here, and the top screenful or two of the _haricrash file. It may output a lot of gibberish. I don’t need all of it! ;-) Just enough to trace what’s actually happening.

I’m just wondering if it’s something else that is acting up here, as Harinezumi v0.09 isn’t as big as the address given (it’s 13016 bytes, the address is at +77212 bytes).

Hopefully the two bits of information given will shed some light on what’s going on.

Thank you.

where information is immediately written to a disc file.

Probably not the best idea if the system is giving (spurious?) broken directory errors.

I don’t believe that Rick has a machine with that much RAM,

I don’t. My Pi2 has 1GiB with habitually ~920MiB remaining unused. So there never seemed much point in going for more.

That said, the RMA is in low memory (&20000000) and the application loads into low memory (&8000), so nothing there ought to be tripping the system up.

Given it’s an undefined instruction, it’s likely that there’s some code sequence (possibly invalid) that worked on earlier machines, but the Pi4’s processor is choking on. Be interesting to see what the instruction is…

You’re running version 0.09, yes?

I suspect v0.07.

1. The web site offered no other
2. Earlier today the web site did NOT (as in not respond, not anything – turned off?)

Bernards problem looks to be something affecting the filesystem access – duff card, duff connector to card, dodgy PSU for system. Martin got him running Reporter and that was picking up fairly random issues with filesystem read/write.

Thanks Rick Murray

Check the cable between the power supply and the Pi. Some can be rubbish and deliver rather less than the desired 5V. If in doubt, change it.

Power at GPIO pins when running is 5.20v.
Power cable built in to PSU and purchased from reputable Raspberry Pi supplier.