RISC OS Direct 5.31 upgrade with Wifi and Web Browser

*vcgencmd bootloader_version
2024/05/17 12:26:58

Hmmm. I’ve just noticed that the Pi 4 EEPROM release notes contain the following entry for the 2024-05-17 EEPROM release:

Add recovery_reboot option to config.txt for rpiboot which causes the system to reboot after updating the bootloader.

Could it be a simple matter of adding the appropriate “recovery_reboot” line to config.txt? Sadly I’ve not been able to find any documentation for this option…

Removing the firmware update files before booting into RISC OS on a RPi 4B worked OK.

Potentially more interesting firmware update.

Introduce a new boot-menu feature where pressing SPACE at power on gives the user a one-shot option to select a different boot mode. e.g. Select USB boot if the default SD card is corrupted or unavailable.

Interesting that he points out BASIC being case sensitive in that KEYWORDS must be in CAPITALS. Isn’t that how most traditional BASICs are, with VisualBasic famously being different with it’s case insensitivity.

The Amstrad CPC’s Locomotive BASIC allowed you to type commands and functions in lower case or mixed case, but tokenised them. When you listed the programme they were then all shown in CAPITALS. It was very convenient.

Back when BASIC was invented, as with FORTRAN, there was only UPPER CASE.

YES, BACK IN THE DAYS WHEN ALL CODE WAS WRITTEN IN SHOUT. ;)

The reason we have so many bugs these days is because we don’t shout at the machine any more. It isn’t scared of us. Additionally, since most modern devices are little black blobs that “do magic”, there are precious few people capable of wielding a test probe in a threatening manner.

Isn’t that how most traditional BASICs are, with VisualBasic famously being different with it’s case insensitivity.

I don’t keep a list of BASIC variants’ idiosyncrasies, but I know that Locomotive BASIC and Mallard BASIC on Amstrad/Schneider CPC and PCW/Joyce did the sensible thing: they were case-insensitive on input and uppercased on output (i.e. LIST), which made detection of typos in commands very very easy.

Everyone I ever spoke with from the 16bit brigade was shocked when faced with the BBC BASIC “uppercase only or else Syntax Error” behaviour (especially the Atari ST guys used to GFA Basic which was really many miles ahead of BBC BASIC). Was there ever a proper BASIC editor on RISC OS that fixed that problem?

I like BASIC the way it is. By typing everything except keywords in lowercase, I avoid accidentally including a keyword in a variable name – which would get messy.

we don’t shout at the machine any more. It isn’t scared of us.

Speak for yourself. :)
Stuff still starts working when I enter the fray, no discernable reason, it just does. ???

All of these “failure to boot” incidents have been with a Pi4B rev 1.5, right?

It may simply be that the EEPROM upgrade supplied with RISC OS Direct v2 (January 2021) is too old for the Pi4B rev 1.5 (December 2021) and the attempt to run recovery.bin fails as a result.

Of course, installing a January 2021 EEPROM may not be a good idea on other Pi 4 units, so it’s best to get rid of the files (pieeprom.upd, recovery.bin etc.) anyway.

I had only cursory experience of RISC OS Direct with the (now severely outdated) RPCEmu bundle, so I decided to try it now as a basis for updating my various Pis (which are mostly on 5.24). My experience so far putting the v2 onto a microSD card and then applying the v3 update. One RPi 4 (1 GiB of RAM) and one RPi 400. There are no contact details I have found on either the “Direct” download page or the accompanying ReadMes, so I just put my feedback here.

• https://www.riscosdev.com/ is currently showing a Wordpress error: “There has been a critical error on this website.”
• the image is slightly too large for all 6 of my empty 16 “GB” microSD cards
• no doc is accompanying the image – it is the only file in the ZIP
• it turns out that the Filecore partition is only (slightly less than) 8 GiB of size, which makes it even worse to choose an image size slightly too large for 16 GB cards
• startup results in 640×48075 Hz - this might be a special case because of my setup (native 4K monitor behind a KVM Switch which only does 4K30Hz), but other OSes have no problem detecting this
• The “Function Keys” thing tells me “F2” is “Load” – really? Which application?
• Why the mouse buttons explanation graphic is called “Mouse Shortcuts”, I don’t know/understand
• when StrongEd is opened, it does so in “a64” column size, which shows the ReadMes with unfortunate line breaks
• the fact that the “5.31 upgrade” needs to be applied from within RISC OS is neatly hidden in the !ReadMe inside the ZIP, which is not easily opened in a non-RISC OS environment
• there are no contact details anywhere
• the upgrade procedure is non-obvious and you need to read a lot to complete it successfully
• the upgrade !ReadMe promises a ready-made image of RO Direct including the content of the update, but this seems to be unavailable
• even after the upgrade, there is still an “IMPORTANT.TXT” file in the FAT boot partition that looks very outdated (probably from Direct v1)
• The Wifi gizmo that shows all the available networks fails to visualize my network names fully, which means I cannot distinguish my guest Wifi from the real Wifi, because they both share a common prefix
• after the upgrade, again back to 640×480@75Hz, the shown “intro to Direct 5.31” window is not fully visible (and therefore the buttons at the bottom are not accessible) because it is too large for this resolution and does not have scrollbars
• the boot stuff tells me that ARMv5 compatibility mode is active – I don’t believe that this can be technically true
• Iris starts up with an arcane (for new users!) error about an (un)aligned memory access setting – maybe it would be a good idea to deliver the whole RISC OS config in a way so that it “just works”?
• when starting StrongEd again and a text file was opened the last time, an error pops up: “Unknown Bitmap”

I am a seasoned RISC OS user, so overcoming these things (and even understanding the history of all these niggles) is easy for me. If I was someone curious about RISC OS with no previous experience, i would have probably just given up. Too many rough edges. Not a user-friendly experience at all (it is much better than “naked RISC OS” of course, but much worse than “any old Linux distrib you can get for your RPi”).

Was there ever a proper BASIC editor on RISC OS that fixed that problem?

StrongED’s comes with an add-on (KeywordFix) that converts BASIC keywords to uppercase. By default this is tied to the Return key so after Return is pressed all keywords on the line will be changed to uppercase.

Written out of frustration of having to stop coding and having to go back to fix small typos such as ‘THEn’. Often not spotted until running the program.

when starting StrongEd again and a text file was opened the last time, an error pops up: “Unknown Bitmap”

I don’t know which version of StrongED is on v2 but the v3 update comes with 4.70a15 which is a test version that I do not consider suitable for a general release.

Also the instructions say to select everything in the update and drop it over the v2 install which is definitely not the way to update StrongED due to its complex internal structure (which is likely to change over time).

@ Steffen
The ROD contact address is info@riscosdev.com
The Direct webpage does recommend a 32GB card
You might be interested to know that Andrew Rawnsley will be talking about the latest version of RISC OS Direct on Wednesday night at the WROCC club meeting via ZOOM.
Iris has always needed unaligned memory access as it uses bits of Thumb 2 which is part of the ported code.

The ROD contact address is info@riscosdev.com

The main webpage does not work – do you think email will work? Anyway, this was not my point.

The Direct webpage does recommend a 32GB card

It does, in the fine print, with the annotation “but some 16GB cards may suffice”, which I dispute with the experience of “0 of 6”.

Using a 32 GB card means that 24 GB will be wasted instead of 8 GB. Not that it makes any difference, since microSD cards got so cheap that it does not really matter if you buy one new. But if all you have currently at hand is an empty 8 or 16 GB card, you will be p*ssed off to find out that the reason these didn’t work was just because of the carelessness of the producer of the image.

Iris has always needed unaligned memory access as it uses bits of Thumb 2 which is part of the ported code.

This might be true, but misses my point. I complain about user unfriendliness. Nobody outside of deeply knowledgeable RISC OS users will understand the error message, and it sounds really frightening. And if you want to provide an out-of-the-box experience, you need to configure the box correctly.

While we are at it, the RO Direct history page also has interesting “More info…” links. To the same page.

The phrase "Versions of RISC OS with an even number are considered “long term suport” builds." disagrees with ROOL’s definition here

I fear that my day-job as the QA guy for mission-critical software might be colouring my view on what kind of quality I should be presented with.

it turns out that the Filecore partition is only (slightly less than) 8 GiB of size, which makes it even worse to choose an image size slightly too large for 16 GB cards

I wrote an image designed for a 16GB SD card to a 32GB SD card and then tinkered with it to get it right. I then saved it back as an image but of course I got back something just over 30GB. I therefore truncated it back to the original image size.

The simple solution here is to use BBC BASIC for Windows to reduce the image size here it is reduced to 14.something GB:

   10 REM >SetSize
   20 :
      F%=OPENUP"C:\Program Files\Win32DI\30G_armini.img"
      PTR#F%=0
      L%=1392508928
      H%=7
      SYS "SetFilePointer",@hfile%(F%),L%,^H%,0
      SYS "SetEndOfFile",@hfile%(F%)
      SYS "CloseHandle",@hfile%(F%)
      PRINT F%
      END

The main webpage does not work

Well thats fixed for you now as its up this morning.

I fear that my day-job as the QA guy for mission-critical software might be colouring my view

Like missing the obvious that the site was down due to a fault the owner was not aware of or perhaps moving servers? :-)

Steffen’s whole point is that the whole thing is not a simple ‘load and go’ solution for new users.

The simple solution here is to use BBC BASIC for Windows

And writing that in BBC BASIC for Windows is probably not simple even for most people here!

Incidentally on Linux you can do it with one line in a terminal

truncate -s <size in bytes> <filename>

truncate -s

Direct v2 is now on an 8GB card after doing this on the Mac

truncate -s 7G test.img

GParted shows the filecore partition to be 6.94GiB and there is still 287GiB of unallocated space to verify that too much has not been chopped off.

(The exact end of the partition can be found with fdisk -l.)

Another potential problem with the download. The zip file format (deflate64) used to compress the image isn’t compatible with RPi Imager so needs to be decompressed first.

Using xz to compress the image saves a bit of space and can be written with RPi Imager directly.

1289650573 Nov 1 13:31 RISCOS_DIRECT_528_Jan_21_a.zip
1132618272 Jan 28 2021 RISCOS_DIRECT_528_Jan_21.img.xz

StrongED’s comes with an add-on (KeywordFix) that converts BASIC keywords to uppercase. By default this is tied to the Return key so after Return is pressed all keywords on the line will be changed to uppercase.

That is useful, thanks for the info. My times as a BASIC programmer are long gone (and I will avoid telling you that I was a Zap user back then…), but nice to know that there is now a solution.

Also the instructions say to select everything in the update and drop it over the v2 install which is definitely not the way to update StrongED due to its complex internal structure (which is likely to change over time).

It also does not mention the “copy” options – Force? Newer? Defaults (whatever they are) may work?

I really like some of the things that accompany the upgrade, like the info screens and the options to set helpful tweaks for your machine like a key for the middle mouse button and a “move F12 to F10 because the RPi 400 designers were braindead”, but there are also many niggles in the upgrade process.

Iris starts up with an arcane (for new users!) error about an (un)aligned memory access setting

It sounds a bit weird for a currently maintained application to be being built for a memory access style which requires poking around in the ARM control registers to reverse the default (safer/stricter) OS setting. For something as complex as a web browser I think I’d want to cling on to any memory protection I could. Why not just compile the code with the appropriate compiler switch?

When I hear justifications “because Thumb2” that doesn’t stack up for my reading of the ARM ARM – Thumb2 instructions must be halfword aligned, so no problem there as that’s also legal to do with alignment exceptions dialled up to the max (just as you can use LDRH/STRH, Thumb2 fetches are similar). Curious…

Having had a poke around with the 5.31 direct features, the following feedback:
- The mp4 player starts relatively quickly, though I can’t reach the actual program in a map. Apparently it’s part of !boot, but I’m not going to fidget around in there for safety. There are keyboard commands that do the usual video stuff (play, pause, forward, backward, ect…) but they aren’t to be seen anywhere in a hints menu, or as buttons at the bottom. With guesswork I was only able to find out the basics. Standard definition videos at around 480p are buttery smooth, higher resolutions make the video lag. The splash screen on startup is very Windows 8, covering up everything behind it with seemingly no way of stopping it.
- Wifi works, I haven’t done speed tests. When the signal is weak and flickers in and out of a connection, it kept asking my password, even though I already have it saved. With the signal stronger, this problems no longer occured.
- Iris (big one):
1) I again haven’t done a speed test, RP4 is in a metal case quite far from the hotspot, so it would not be fair to. An image heavy site like “Knowyourmeme.com” loaded, seemingly with no problem. When I clicked an image, first an error message appears (I think something about something missing). After going to the previous page and back to the image, it loaded no problem. If you want (or someone) I can post the error message here.
2) Next came the Google test. I searched for an old website for funsies: Space Jam 1996. The search results appeared, but every time, when the bar reached +/- 95%, everything locked up and had to hard reset my PI. No idea what could be causing it.
3) Youtube is a blank page. This was probably a given, maybe some day.
4) Outlook email: I had the reflex to right-click in a sub folder, to select “mark all as read”. I presume the browser prioritised Risc OS context for right-click, as no menu appeared. How does a request for a menu work inside the browser, on a site that expects right-click to create a menu screen? The middle-click made the Iris menu appear, which was according to my expectations.
5) There was talk about some hardware acceleration during a Risc OS dev talk, which could help Iris?
6) Iris also creates a splash screen on start-up, covering everything behind it.

Thanks for the update. Will be playing around some more when there’s time.

It sounds a bit weird for a currently maintained application to be being built for a memory access style which requires poking around in the ARM control registers to reverse the default (safer/stricter) OS setting. For something as complex as a web browser I think I’d want to cling on to any memory protection I could. Why not just compile the code with the appropriate compiler switch?

It’s likely that WebKit is assuming that ARMv7 and later always has unaligned memory accesses available – IIRC that’s true on most other armhf and aarch64 platforms I’ve encountered (and on x86), and some programs and libraries have specific optimisations when unaligned access is known to be available, so it’s quite possible that ARMv7 with alignment exceptions enabled is something that the upstream WebKit developers haven’t been interested in supporting for ARM-specific code.