BB -xm ROM 29/03.13

Updating from 25/3/13

Before getting to the Castle logo I get the error ‘RAMFS: Filer may have gone wrong’. Description offers ‘Not a heap block’. Machine hangs.

I configured no RAM disc (reverted ROM to do so). This time at the same stage the error given was ‘SDFS: Filer may have gone wrong’. Same description. I couldn’t use Escape during Boot to stop !Boot to switch ROM images this time obviously so just as well I have a spare uSD loaded.

BB -xm Rev C1

I have the same issues with the 29/03/2013 OMAP4 ROM on the PandaBoard-ES

got the same problem on the pandora with the
rom from 29.03.2013

Also true for 30/03/13 ROM.

Pressing [Escape] during start up followed by *Desktop gets a desktop without RamFS or SCSI disc icons, only SDFS being available. Sometimes [Escape] just gets a data abort error & hangs.

It looks like this is being caused by some recent changes to sprite handling in the Wimp, as reverting to WindowManager 5.19 seems to make the problems go away. But I’m not sure yet what the actual bug is. After trying every machine I have except the TouchBook (maybe tomorrow), the only one I’ve been able to reproduce it on is the Pandora, which isn’t exactly the easiest to use for debugging.

If it is the Wimp changes which are causing the problems (which it looks like it could be, as I also saw the Desktop module fall over on the Pandora due to it improperly handling an error the Wimp was throwing), then perhaps whether it triggers or not is entirely dependent on what types of icons/sprites your programs use on startup. Out of the machines I’ve tried, the Pandora is the only one which will have BatMgr running, so maybe that’s triggering it. If that’s the case I should be able to build a test version of the module and debug the issue using a BB instead of the Pandora.

It looks like this is being caused by some recent changes to sprite handling in the Wimp, as reverting to WindowManager 5.19 seems to make the problems go away. But I’m not sure yet what the actual bug is. After trying every machine I have except the TouchBook (maybe tomorrow), the only one I’ve been able to reproduce it on is the Pandora, which isn’t exactly the easiest to use for debugging.

Try RPCEmu. The 26th March ROM is OK, both 29th and 30th error at the desktop stage.
As I said elsewhere – RPCEmu 0.8.10 definitely fails, I need to do a couple of twiddles to revert to 0.8.9 to check that also fails in all circumstances.

Edit.
Yep, running either the 0.8.9 recompiler (or interpreter) version also causes the desktop errors.
The nice thing about RPCEmu, is it’s rather easy to swap stuff.

RPCEmu 0.8.10 definitely fails, I need to do a couple of twiddles to revert to 0.8.9 to check that also fails in all circumstances.

31-Mar-2013 works for me on RPCEmu 0.8.9.

31-Mar-2013 works for me on RPCEmu 0.8.9

One has only just finished feeding the small people in fur suits¹ for the second time, downloading the latest ROM is on the todo list.

¹ Commonly known as cats. Also serve as alarm clocks permanently fixed on GMT (Great Meal Time)

I have the same issues with the 31-Mar-2013 OMAP4 ROM on the PandaBoard-ES

Still having the same issue with 31/03/13 on my BB -xm.

Out of the machines I’ve tried, the Pandora is the only one which will have BatMgr running, so maybe that’s triggering it.

On the BB -xm BatMgr errors during mod init, which presumably it should. So…?

Still having the same issue with 31/03/13 on my BB -xm.

I’m so glad it isn’t just me.

@Sprow Is that just a basic emulator install and the HD4 boot unpacked? i.e. bog basic machine waiting for other stuff. What version boot?

Sprow Is that just a basic emulator install and the HD4 boot unpacked?

Yep. HostFS with HardDisc4.util installed. !Boot is 1.11 from 08-Oct-12. I even cleaned out my Choices directory.

Yep. HostFS with HardDisc4.util installed. !Boot is 1.11 from 08-Oct-12. I even cleaned out my Choices directory.

OK. Rip down time. Then re-test.

Out of the machines I’ve tried, the Pandora is the only one which will have BatMgr running, so maybe that’s triggering it.

On the BB -xm BatMgr errors during mod init, which presumably it should. So…?

That’s normal – on everything other than a Pandora or TouchBook BatMgr will throw an error on init and refuse to start.

My BB-xM (with modified BatMgr) and TouchBook seem to be immune to the bug, so I’ve been focusing on debugging it using the Pandora. It looks like ColourTrans is straight-up lying about how large some translation tables are, causing it to overrun the buffer the Wimp allocates and corrupt the heap. I’ve tracked it down to the exact call that’s causing the problem, so hopefully it won’t be too long until I find where it’s going wrong.

My BB-xM (with modified BatMgr) and TouchBook seem to be immune to the bug, so I’ve been focusing on debugging it using the Pandora.

Following the comment from Sprow I ripped out (well renamed) the old hostfs and installed a clean HD4.
Errors repeatedly until I renamed the cmos file and made it create a new one. A few reboots along the way and it booted fine. I’m wondering if the cmos file in RPCEmu was holding something that the new build didn’t like.

Ah, just tested the setup forcing the system to 800 × 600 with 16million colour and regenerated the boot errors. In fact 32k colours will produce the error. For people who want to play the quick escape is setting the emulator to no VRAM and thus force a 256 colour mode – 800 × 600 256 is OK.

I’ve now submitted a fix. The problem was down to the subtle differences between ColourTrans_SelectTable and ColourTrans_GenerateTable – ColourTrans_GenerateTable was being used to generate the table (with bit 4 of R5 set to request a wide table to be generated where applicable), but ColourTrans_SelectTable (which sometimes ignores R5 entirely) was being used to read the table size. So depending on the source sprite & dest screen mode, SelectTable would sometimes return an inaccurate size for the later GenerateTable call.

I’ve now submitted a fix

Sorry Jeffrey, but I still have a problem with the latest ROM image. I don’t know whether it’s connected to the bug that you’ve just fixed, but the latest ROM images seem to have disabled the Delete and Backspace keys on my Beagleboard xM, when using Ovation Pro. Reverting to 17 March ROM works. Interestingly, while writing this note, I was unable to use the Delete key even though the I’m using the 17th March ROM.

I’ve now submitted a fix.

Working fine Jeffrey, RPCEmu shows no boot errors with either 32k or 16M colour.

@Kevin – as I recall¹ Ov Pro has at least one applet that controls the key actions, particularly the PC vs. Acorn style delete actions. The applet may be having difficulties with the new behaviour. You may even have an old applet that you missed an upgrade for.

¹ I don’t use OvP on anything other than the old machine that also does the mail. It doesn’t run RO5 as mucking about with the mail system might stop the catering facility…

@Steve – I’ve checked it out, and the situation is worse than I thought. It’s definitely not an OPro problem. StrongEd4.69a7 and Easiwriter9.01 don’t respond to ANY keyboard presses, while with OPro it just seems to be the Delete and Backspace keys. One thing I have noticed, is that there is a warning reference about Pandora keyboard as the machine starts, but before the boot sequence starts. It passes too quickly for me to read exactly what it says. The same warning is NOT there with the 17 March ROM

Also working here Jeffrey.

@Kevin – What machine? Ovation Pro 2.77u seems to be working fine for me on my BB -xm on a brief test. Deletes OK.

Haven’t got StrongEd or Easiwriter to try.

One thing I have noticed, is that there is a warning reference about Pandora keyboard as the machine starts

It’s probably warning that the Pandora keyboard isn’t present which, AFAIK, is fine so long as you aren’t using a Pandora. There’s a few warnings about absent or irrelevant items during the module initialisation.

Haven’t got StrongEd or Easiwriter to try.

Well the first is a free download
http://stronged.iconbar.com/

Well the first is a free download

Well, running it and entering/deleting text seems OK with the current ROM but that’s just downloading StrongED and using it virgin ‘cos I don’t know the S/Ware. Can’t try forward deletes – my keyboard doesn’t have one.

@Frederick – it’s a beagleboard xM. However, I’m also encountering similar problems with the latest ROM when soft loaded on my Iyonix

I’m also encountering similar problems …. on my Iyonix

Sorry, but that is wrong. It’s working okay on the Iyonix. The BB xM is still not playing, in spite of changing to a back up boot, and downloading today’s ROM again, in case the first download had been corrupted in some way. Everything works fine with the 17 March ROM. Will investigate further.

Just tried the latest ROM 2013-04-02. Still no joy. Also downloaded and tried a vanilla boot. Keyboard seems unusable. :-( Everything fine if I return to 17 March build.