InetRes updating automatic

How about: Select-click on disc opens the filer with hidden items hidden, Adjust-click opens the filer with them un-hidden?

If !Boot is hidden, really hidden, so that all the stuff that’s Filer_Booted isn’t, then lots of things are going to go wrong. As the evidence shows.

What is displayed or not in a Filer window is irrelevant to commands – they just use the filing systems themselves. In my test everything booted and started as I would expect – my problems started when I used Filer windows.

Presumably hidden items are still booted and set up their variables as normal?

Both !Choices and !Scrap need to be in a known location (the root of the boot drive) so !Boot can find them programmatically.

But…able to be placed back inside Boot for those who understand the desire to separate system resources and user resources, but at the same time don’t want crap cluttering up $.

Absolutely. We need to continue to support the ‘everything tucked inside !Boot’ model because the boot sequence still needs to work on earlier versions of the OS, so the ability to move things outside of !Boot is to help with things like backups (where you don’t want to waste time backing up !Scrap for example) and generally reducing the chance of newcomers getting trapped in quick sand by messing inside the OS.

To avoid a Filer_Visibilty error on starting up with an unmodified ROM the softload will be placed in PreDesk.

As with all the commands in PreDesk they match the OS which was running at the time, but softloading may invalidate that assumption. I did fix an issue with Filer_DClickHold which could be written out by the plug-in despite the host OS not supporting it. Oops! Nobody noticed that for 12 years.
If you re-run the plug-in it’ll parse but ignore the new commands and write out a valid set for whatever ROM you booted.

Trying to do much with the Filer seems to abort, which makes debugging awkward! Opening a directory gives Data Transfer abort in Filer @&FC25F58C +70A8, which is an LDRB,[r1},#1 but r1 is &6554006B.

Thanks for the report, sounds like there’s probably a bug there – I did have to shuffle a few registers around to free one up and looks like it’s picked up a few letters of a string and tried to use them as a pointer. Maybe hold off testing the hiding option until I figure out what’s causing that?

Seem to remember that lack of !Sprites in ! directories caused some issues before so is this the case for !Choices?

I think you’re remembering that *IconSprites reports a missing sprites file correctly now, but you’re still allowed to have an application without any sprites.

I have now tried Netsurf 3.11 and it is still the same when system applications are hidden.
StrongEd and Packman seem to play nicely though but Store is another application that doesn’t like the system applications being hidden.

Nothing should care about things being hidden, that’s purely a display thing, the files/directories are still there. Moving !Scrap shouldn’t upset anything since it gets renamed for network boot or people move it to RamFS, but moving Choices might trip up some badly behaved apps. So: try NetSurf with Choices moved out of !Boot, but not hidden.

Presumably hidden items are still booted and set up their variables as normal?

Correct. Aside from that data abort already noted, the hiding is just about not putting the item(s) in the Filer window, they are otherwise run and processed as normal during booting.

Adjust-click opens the filer with them un-hidden?

Uh, doesn’t FAT32FS already have specific behaviour for adjust clicking on the drive icon?

I’m not sure I understand the aversion to adjust click on Switcher anyway.
To me it makes perfect logical sense.
Select on the icon is for managing transient system resources (what apps are running, how memory is allocated, RAMdisc presence/size, etc etc) while Adjust on the icon is for managing permanent system resources (eg fonts, stuff seen at Boot, configuration, etc etc).

Another item in !Boot that would be better outside is Resources:!Manuals
It does not cause a problem with backup as it rarely changes but it makes searching for a string via !Locate a pain as it generates loads of false positives when searching in !Boot for a problem.

Uh, doesn’t FAT32FS already have specific behaviour for adjust clicking on the drive icon?

Personally I would never consider using FAT/FAT32 for a boot drive – or any other ‘internal’ RISC OS drive for that matter. I only use it where I need to transfer between systems and the network isn’t available for whatever reason using ‘external’ drives.

This conversation has long ago gone way off topic.

Please start one or more other topics.

Would be nice if we could split topics…

So: try NetSurf with Choices moved out of !Boot, but not hidden.

OK so !Scrap and !Choices ouside of !Boot and not hidden and !Netsurf works.

!Scrap outside of !Boot and Choices inside !Boot and not hidden and !Netsurf works.

!Scrap outside of !Boot and Choices inside !Boot and system applications hidden and !Netsurf works.

!Scrap and !Choices outside of !Boot and system applications hidden and get the errors.

!Boot loading is really easy to break as I accidentially moved !System out by mistake on one test and on rebooting it just errored and left me at the supervisor prompt.

Thats where Rick application Harinezumi helps as it allows the boot to carry on in most cases and logs the errors.

Really think something like that should be in the standard build.

Trying to do much with the Filer seems to abort, which makes debugging awkward! Opening a directory gives Data Transfer abort in Filer @&FC25F58C +70A8, which is an LDRB,[r1},#1 but r1 is &6554006B.

Thanks for the report, sounds like there’s probably a bug there

…and…

!Scrap and !Choices outside of !Boot and system applications hidden and get the errors.

Slight out-by-20 error, sorry. An amended version should fix the data abort when OS apps are being hidden.

I accidentially moved !System out by mistake on one test

Ouch – a good example of why keeping things the OS relies on, like !System, out of the way of casual shift-dragging is useful too (you wouldn’t go randomly moving files from C:\Windows\System32\drivers I bet!).

And indeed HideSys2 does not abort when hidden, and Filer displays hide things as expected. Will test further … but probably not until Sunday.

And indeed HideSys2 does not abort when hidden

+1

Netsurf now behaves when !Scrap and Choices are removed from !Boot and system application directories hidden, as does !Store.

HideSys2 does not abort when hidden

+=1

All good on the Titanium.

What goes around comes around …

My proposed strategy for the other bits is to try to make !Boot be able to have fewer user things …

When the Universal Boot first appeared – was that about 1996? – there was a reaction against it much along the lines now being echoed here. That was my reaction, too.

The stuff inside of !Boot is more tricky because there’s a web of dependencies …

You’re telling me! At the very beginning of ROL they released a Draw diagram of how everything inside !Boot was linked. It was not even readable unless at A3 size. How much more complex it must be now. ;-(

Ouch – a good example of why keeping things the OS relies on, like !System, out of the way of casual shift-dragging is useful too …

Yet for many years the tens-of-thousands of users in schools happily got on with !System, !Scrap and !SharedClib in the root directory, often along with a handful of others, such as !Fonts.

In those day the modular approach of RISC OS extended to such resources, hence the !\<resource_directory\> naming still found unnecessarily within !Boot. Remember that you can Run a directory without the ‘!’.

happily got on with !System, !Scrap and !SharedClib in the root directory,

I put up with it when using floppies.

When it came to my first harddisc, I sussed that the OS would run an application called !Boot, so I put the resources in there and did some extra setup along the way (Filer_Boot a bunch of things).

I held out against Universal Boot until I got a RiscPC, basically because I found it rather clumsy compared to my setup…which, granted, used hardcoded paths, but still, UniBoot seemed to be really overthinking the process.

When the Universal Boot first appeared – was that about 1996?

!Boot as an umbrella for !System, !Scrap, networking, choices and the boot sequence was “invented” in 1994 with RISC OS 3.5 and the Risc PC. “Universal Boot” was born when the whole Toolbox/Nested WIMP for RISC OS 3.1 was necessary for Browse and Java (1997 I think, post-StrongARM and RISC OS 3.7). First public version of it on the Clan Beta CDs Browse and Java IIRC.

HideSys2 works on the Titanium except that I can’t use three of Alex Waugh’s configure plugins.
‘OpenChoice’, ‘OpenPre’ and ‘OpenTasks’. ‘OpenRes’ works and i think that the other three will be okay with a slight modification to their Run file.
Also of concern (to me least ways) is the fact that I cannot save a new mouse colour via the excellent ‘Pointer Colour’ plugin by Joe Taylor. Very useful for having a mouse pointer coloured to visibly show up against a background. This plugin is in !Boot.RO500.Hook.Res.Configure, placed there by Joe’s program.
As to the RPi4, I am cloning a new SD card to try as I’ve managed to totally ruin the card I was testing with HideSys1. I will test The RPi4 as soon as !CloneDisc has finished its work.
HideSys1 during testing gave me an &FC2272DC error on the Titanium.

Regards Ron.

Correct. Aside from that data abort already noted, the hiding is just about not putting the item(s) in the Filer window, they are otherwise run and processed as normal during booting.

What happens if you use OS_GBPB to enumerate the contents of a directory? For example OS_GBPB 10, 11 or 12: file attributes (bit 6 is undefined so could be used to flag ‘hidden’) and object type (1=file, 2=directory, 3=imagefile)?

I have now tested HideSys2 on the RPi4 and all is well. As I thought Alex Waugh’s configure plugins only needed a change in the run files. As for Joe Taylor’s PtrClr, it is wierd. Changing the Pointer colours and clicking save results in a ‘Not Found Error’, however changing the pointer colours gives a result that persists even after a reboot despite not saving the new colours.

Regards Ron.

file attributes (bit 6 is undefined so could be used to flag ‘hidden’)

Just in case this was not generally known, Filer+ used that bit for that purpose and it is documented in the OS StrongHelp manual.

Here, too.

https://www.riscosopen.org/wiki/documentation/show/FileSwitch%20Key%20Features#file_attributes

My problems with Alex Waugh’s configure plugins and Joe Taylor’s PtrCfg plugin have been solved by the estimable Sprow. All work as intended now.

Regards Ron.

My problems with Alex Waugh’s configure plugins and Joe Taylor’s PtrCfg plugin have been solved by the estimable Sprow. All work as intended now.

For anyone curious, with PtrCfg it was referring to Boot:Choices.Boot.PreDesk contrary to the note in PRM5a-540 which urges you to refer to PreDesk “as <Boot$ToBeLoaded> to preserve for future compatibility, should the structure of boot applications change” or <Choices$Write>.Boot.PreDesk alternatively.

Alex Waugh’s configure plugins used the right system variables in the obey files for RISC OS 4, but weirdly didn’t for RISC OS 5. Merely copying the RISC OS 4 obey files was all that was needed.

Aside from that data abort already noted, the hiding is just about not putting the item(s) in the Filer window, they are otherwise run and processed as normal during booting.

What happens if you use OS_GBPB to enumerate the contents of a directory? For example OS_GBPB 10, 11 or 12: file attributes (bit 6 is undefined so could be used to flag ‘hidden’) and object type (1=file, 2=directory, 3=imagefile)?

Hiding is just about not putting the item(s) in the Filer window, they are otherwise run and processed as normal by OS_GBPB et al.

I realised the Filer might be better faulting attempts to copy or move an unhidden system application (eg. on an install floppy) to the root of the boot drive as it would appear to vanish into a black hole, so will have a think about how best to present that.