Partition Manager

I don’t think I’m going to get away without a bespoke AcornURL Module that doesn’t have the !Boot dependency

Having spent way too many hours try to break the dependency on !Boot, I don’t believe its possible without AcornHTTP/AcornSSL/URL_Fetcher being modified to either cache the files they use or copy them into Resources. I’ve been staring at their code, trying to figure out what it wants from !Boot, but can’t see anything obvious. It’s the call to URL_GetURL to initiate the transfer that’s triggering it, so possibly Protocol related, but it’s not obvious from the source code

Worse still, I might have to prevent the drive/filesystem hosting !Boot from being formatted as I’m seeing several fatal crashes within SDFS and FileCore after it requests the original SD be reinserted. I need to investigate these further and report them s bugs, one may be related to a bug I reported last week as I’m seeing an invalid error pointer returned after it gives up looking for the original SD, but I’m not sure if the other Aborts are knock-on effects of the initial error or additional bugs.

There’s a “final” v0.99 available for any last minute bug reports, before I bump it to v1.00 and call it “stable”. I’ll package it at that point and see if ROOL would like to replace !HForm with it and remove that requirement from the Filing system improvements (step 2) bounty

Should I BasCompress the final release? I’ve modified the way it uses LIBRARY and OVERLAY to work around an issue with BasCompress (no support for external LIBRARY files) and from a local test it looks like it works – I’m in two minds about it needing separate testing. I would really like to get my hands on the source of BasCompress to fix the issues I’m seeing though, as I’m having to manually fix a few things after compressing it. BasCompress also can’t remove the debug entries, which add quite a lot to the size to it – another manual step when compressing.

Please include the uncompressed source BASIC file in a sub directory inside the app.

Echoing what I said back in February, there is currently no licence included and therefore usage and distribution rights are unknown. Please add one.

Having spent way too many hours try to break the dependency on !Boot, I don’t believe its possible without AcornHTTP/AcornSSL/URL_Fetcher being modified to either cache the files they use or copy them into Resources. I’ve been staring at their code, trying to figure out what it wants from !Boot, but can’t see anything obvious. It’s the call to URL_GetURL to initiate the transfer that’s triggering it, so possibly Protocol related, but it’s not obvious from the source code

Just a guess – is it the HTTP module trying to read its cookie file from disc? https://gitlab.riscosopen.org/RiscOS/Sources/Networking/Fetchers/HTTP/-/blob/master/c/config – look at the function config_open_cookies_for_read which in turn is calling config_try_open which reads from CHOICES_DIR (defined as “<Choices$Write>.WWW” in the header file). I think if you try to open a file on a disc that has been removed you should be prompted to reinsert it – this might be a bug in HTTP.

Should I BasCompress the final release? I’ve modified the way it uses LIBRARY and OVERLAY to work around an issue with BasCompress (no support for external LIBRARY files)

You might want to look at Crunchie by the late Steve Drain although I’m not sure it handles OVERLAY libraries. http://kappa.me.uk/download.htm

Regarding crunching the BASIC… why bother? What is the point?

Re. ’There’s a final . . for any last minute bug reports: you asked for it so, 3 bugreports:

!Runimage line 305 one unneeded? quotation mark (‘"’) after ‘“_SectorOp”’ in PROCdebug_text. Line 304 has a similar call and ‘_SectorDiscOp’ does not have this ‘"’
!Runimage line 377 Missing closing ‘)’ for PROCinformation
Formats.Filecore line 775 two ‘((’ too many in statement Boot%!&1F8=((Partition_info%(D%, P%, P_Size_In_LBA_Low%)
FS.ZIDEFS line 289 this ‘ENDIF’ probably unneeded (unimportant).

i really hate to write this down because i admire this work of yours (and ADFFS as well). But small errors can also crash a program, they fare better without.
I would not crunch the program. !PartMgr appears to be a very complicated application, that needs to be 100% trustworthy. A cruncher that encounters a never-yet-seen construct can easily introduce errors in the crunched product. You can save around half the space (from >300K to 150K) but it is not worth the risks. For example, ‘resolution$’ and ‘option$’ appear in DATA lines and must be excluded from crunching. And also a lot of functionnames that are EVALuated must be left as they are.

3 bugreports

Thanks, that’s just the sort of feedback I need. I’ve uploaded a corrected version

I would not crunch the program

My reasoning behind compressing it would be if it was going into the ROM. I’ve not offered a compressed version for download for the very reasons you’ve stated – its not worth the risk when you’re potentially going to give someone a very bad day. It would need a whole new round of testing.

I have worked out the BCspecial config file, with functions left unchanged for the EVAL’s. In reality it’s only removing whitespace, REM’s and reducing variable names. To remove all the debug code, I’d have to search/replace in StrongEd as BasCompress doesn’t support anything like RegEx replace.

Three “oddities” noticed with Partition Manager 1.00 (23 July 2023).

Hardware: Raspberry Pi 4B (4GB)
OS: RISC OS 5.28 (December 2020)

1. PM tries to close files on wrong SCSI drive
   When using Partition Manager with two SCSI drives mounted (e.g. an SSD as drive 4 and an SD adapter as drive 0), any attempt to modify drive 0 will, at the end of the process, try to dismount drive 4.
   This produces an error message “There are files on the SCSI filesystem that need to be forcibly closed. Are you sure you want to proceed?”.
   Clicking on “Yes” results in all SCSI drive icons disappearing from the desktop (Ctrl-Break needed).
   Clicking on “No” shows that the intended action (e.g. add a FAT partition to drive 0) has in fact worked.
2. Cannot see devices on “4 in 1” SD card readers
   When using a “4 in 1” SD card reader (i.e. one that displays multiple drive icons on the desktop) PM can only “see” cards on the first drive icon (typically “:0”). Any card inserted into other drives (e.g. “:1”) will not be detected by PM.
   I suspect this is a case of “PM doesn’t step past empty drives” rather than a feature unique to the card reader(s) thus affected.
3. macOS partitions always give a “size” warning
   For drives initialised on macOS Ventura 13.4.1, PM will report “Partition 1 on SCSI 0:1 exceeds the drive size, proceed with caution”. The partition size displayed by PM will usually be larger than the drive capacity displayed by PM (but not always: the two displayed numbers will occasionally be equal).

Thank you for the feedback. Could you please attach debug logs (hold ALT when you launch PM) with the issues reproduced.

PM tries to close files on wrong SCSI drive

When a SCSI drive is modified, PM first attempts to use Service_SCSIDetached / Service_SCSIAttached and if they fail, the fall-back is to restarting the filesystem Module and its filer Module. In your case it sounds like the service route failed, so it scanned for open files on all SCSI drives and reported, before attempting to restart the whole filesystem via the following steps:

1. Issue OS_Args 255 to all open files to ensure they’re flushed to disc
2. RMReinit SCSIFS
3. Send a Wimp message to SCSIFiler asking it to Quit
4. RMReinit SCSIFiler
5. Repeatedly send Service_StartFiler, followed by Wimp_StartTask with the response until it receives a response from SCSIFiler

One of the last 3 steps has failed in your case – the debug log might help diagnose which step.

Cannot see devices on “4 in 1” SD card readers

I believe they’re added as SCSI drives? PM scans all 32 SCSI ID’s (4 cards * 8 ID’s), so shouldn’t miss any drives. Again a debug log might help diagnose.

macOS partitions always give a “size” warning

The debug log will confirm, but it sounds like macOS has created a partition that exceeds the drive limits. Its possible its an “out by 1” error on the macOS side – look at the debug log to confirm the exact partition and drive sizes in LBA. If it’s more than 1 sector out, it probably needs reporting to Apple.

I can confirm the first two ‘oddities’ as reported by Stuart: the disappearence of the scsi drive icons and the failure of detection of cards in een SD card reader.

PM does recognise USB sticks, but initialising them as a Pi Boot drive fails. While writing the partition table, it asks to insert the very stick. Either clicking Cancel, or re-inserting the stick stops the process, leaving me with an empty Loader file and nothing else. The debug file is about 155K – where should I send it to?

For me as a total ignorant, a good !Help file would actually be nice. The program as it is now, assumes quite some knowledge.

I did manage however to re-initialise the scsi SSD in my 4té2 as a RISC OS one. The Pi Imager is still there, visible when booting without a SD card, but is it (or will it be) possible to create two partitions so that half of the SSD is RISC OS and the other half is Linux? – which actually is how the machine is configured when delivered…

While writing the partition table, it asks to insert the very stick

I’ve only seen that if there’s a drive name conflict, or when attempting to format the filesystem the OS boot drive is on. It’s listed in the Known issues/caveats section on the download page.

where should I send it to?

See !Help if you need to eMail logs.

For me as a total ignorant, a good !Help file would actually be nice. The program as it is now, assumes quite some knowledge.

Some basic understanding of drives and drive partitioning is required to use PM. I certainly would not advocate anyone use it that isn’t already familiar with drive partitioning tools.

Although I don’t plan to add much to !Help, I will probably create a “How to” web article to cover the Pi Boot drive option at some point.

I did manage however to re-initialise the scsi SSD in my 4té2 as a RISC OS one. The Pi Imager is still there, visible when booting without a SD card

I’m not sure what you mean. Are you saying you initialised the drive as “FileCore” and then formatted it as a full-disc FileCore drive, or you used “Pi Boot Drive” to initialise the SSD for a Pi?

is it (or will it be) possible to create two partitions so that half of the SSD is RISC OS and the other half is Linux? – which actually is how the machine is configured when delivered…

Possibly. You’d have to initialise the drive as whatever partition method the 4té2 OS understands, then create a partition that’s half the drive size formatted as FileCore. Then move the drive over to Linux and add another partition formatted for Linux.

That said, RISC OS doesn’t (currently) understand partitioning, so unless the 4té2 has a bespoke implementation of SCSIFS or ADFS4 that does support partitioning you’re in the realms of using MBR partitioning with an overlayed FileCore drive – along the lines of the Pi.

If it’s the later, you could try initialising as a “Pi Boot drive” and reduce the RISC OS “Size in MB” to 50% of the drive, then move it over to Linux and add the Linux partition with GParted or similar.

Thank you, Jon. Congratulations on the release. I look forward to giving it a spin, as I spotted it on PackMan this morning. :)

Could you please attach debug logs (hold ALT when you launch PM) with the issues reproduced.

Debug logs sent by email.

Debug logs sent by email.

Thanks, I’ve been through them…

PM tries to close files on wrong SCSI drive

It looks like I’ve been overly cautious when detaching logical drives and am initially checking the whole filesystem. I’ll modify it to only check for open files on the drive being detached.

Cannot see devices on “4 in 1” SD card readers

What drives am I expecting to see? One SCSI drive responded as SCSI ID 0:0, which is SCSI::SSD4. The only other drive is your boot SD.

macOS partitions always give a “size” warning

Unless my understanding of exclusive/inclusive is wrong for drive sizes and MBA partition sizes, it is an “out by 1” error. The drive size is reported as &1CF7FFF (inclusive) and the partition starts at &800 size &1CF7800 (inclusive), so ends at LBA &1CF8000 (inclusive) – so is 1 LBA beyond the drive size.

I’ll alter the check as it’s not likely to cause an issue.

I’m sure I remember reading years ago that this was a remarkably common off-by-one error – the higher number should be the highest block number, not the number of blocks.

What drives am I expecting to see? One SCSI drive responded as SCSI ID 0:0, which is SCSI::SSD4. The only other drive is your boot SD.

*scsidevices
Device Type              Capacity    Vendor   Product          Revision
0:0.0  Direct-access      112 Gbytes ASMT     2105             0   
0:1.0  Direct-access     Unknown     Generic  MassStorageClass 1539
0:1.1  Direct-access       30 Gbytes Generic  MassStorageClass 1539

SCSI device 0:0.0 appears on the icon bar as SCSI drive :4, as you mentioned.
SCSI device 0:1.0 appears on the icon bar as SCSI drive :0. The slot is empty.
SCSI device 0:1.1 appears on the icon bar as SCSI drive :1. It has a 32GB microSD card inserted.

Cannot see devices on “4 in 1” SD card readers

Could you see if v1.01 picks up the missing device. You can post the debug log to that thread if required.

PM tries to close files on wrong SCSI drive
macOS partitions always give a “size” warning

Hopefully these issues are also resolved with 1.01

Could you see if v1.01 picks up the missing device.

My Hama SD>USB reader is still not recognised.

Cannot see devices on “4 in 1” SD card readers /
My Hama SD>USB reader is still not recognised.

I mentioned this on the JASPP feedback thread, but where I’ve added a SCSI REPORT LUNs command to look for additional LUN’s; it doesn’t look like its being translated to a USB Get Max LUN command on a Mass Storage Class device as it errors.

Looking at the debug log, it also looks like SCSI_Initialise 2 to LUN 0 on the device is returning an error when there’s no card in the slot – even though the LUN technically exists regardless of a card being in the slot. *SCSIDevices appears to show LUN 0 correctly, so I’ll probably get one of you to try some SCSI SWI to see what the error is – its possible its an undocumented error that PM needs to ignore.

Clicking on “Yes” results in all SCSI drive icons disappearing from the desktop (Ctrl-Break needed).

Hopefully today’s build will resolve that issue. I tracked the related Abort you were seeing to an unaligned read in SCSIFiler’s HotPlug UpCall handler – I’ve yet to figure out if it is a bug or something PM is doing though.

The SOE that was causing the UpCall handler to be called in the first place should not have occurred under RO5, which was an issue in the PM Code which is now corrected.

Ooh, now there’s something, and someone might know the answer to a question I had.

Unrelated to Partition Manager itself, I notice some of the screenshots have GPT disks in them, are GPT USB sticks supposed to work on RISC OS?

are GPT USB sticks supposed to work on RISC OS?

No, that is still work-in-progress.

My Hama SD>USB reader is still not recognised.

Could you see if this build resolves the issue please.

YES, now the stick is seen. Do you need a log file?

Do you need a log file?

Not for that, but could you please see if the issue you had when creating a Pi Boot drive is also resolved with that build and if not, send me a log file.