ROMS, ROMS, ROMS

Yes I’m running the 22nd Dec ROM, and get the same thing, I’ve run several tests, and as yet only !Grapevine freezes the machine, I’ll probably revert back to the ROM from the 15th Dec, as I need to access IRC.

I’m going to try the latest builds on RPCEmu and the Iyonix, and see if it’s just an issue with the OMAP3 build.

Curiouser and curiouser! ;-(

Having used the 22 Dec ROM successfully, and having nothing better to do, I decided to load Grapevine to see what happened. Sure enough, it stiffed the machine. However, since then, using the same ROM, I’ve been unable to use Netfetch without the machine freezing, and this in spite of removing Grapevine from the machine!!! As before, returning to Rob’s 15 Dec ROM seems to cure the problem, though I’ve not dared to use Grapevine again.

I’m getting the same results on the Iyonix as on the BeagleBoard, an older ROM build is fine, but using an up to date ROM causes !Grapevine to freeze.

I’ve not had a chance to test out RPCEmu yet.

Right, I think I’ve found the issue that is causing the freeze.
This ROM build is working for me, and !Grapevine is working as it should!

Can you guys please test it out.

23/12/2011 – OMAP3 ROM Download

Rob

Tried your latest, 23/12/11, ROM and !Grapevine works OK here as does !Netsurf and Hermes 3.56.

Well done.

What have you found that is causing the issues?

Doug

Glad it works!!
It looks like the problem is caused by this change – https://www.riscosopen.org/viewer/revisions/logs?ident=1323987595-661104.html (On the 15th December)

The ROM I built is bang up to date minus those changes.
(I’ve sent an Email to the author of those changes, informing him of the bug)

Rob,

Thanks for the specifics.

I’ll keep the latest ROM running to test it.

Now just need to get the Hard Disc !Boot image sorted and we should have the latest issues resolved.
Doug

Can you guys please test it out.

I’ve just been using it for a session browsing with NetSurf (r13272), but after a few minutes NS suddenly crashed; immediately after that I had “Filecore in use”. That’s something I’ve never seen except when using the Zip drive – and it wasn’t even plugged in tonight.

I’m going to try some more browsing with NS r13332 and see if it’s any different. I can’t exclude the possibility that NS is crashing, of course.

It’s happened again. Same version of NS on Iyonix and same site: no problem.

I don’t think the problem is fixed.

But reverting to a build I made on September 22nd (and have been using ever since) shows the same crash.

And I’m happily continuing to browse through many pages of the same site with the same version of NS, on the Iyonix.

Don’t you just hate it when that happens? Symptoms that you can’t rationalise?

Could be a tricky one to sort that Dave, I wonder if the issues with NetSurf and your Zip drive are related?

Have you tried out a clean !Boot sequence?

I suspect I’m going to have to do some rebuilding. It’s as if the first crash has broken something on disc.

I wonder if the issues with NetSurf and your Zip drive are related?

Since the Iyonix and BB behave identically w.r.t. the Zip drive, but completely differently w.r.t. NetSurf, I can’t see how that could possibly be true.

Dave

Been surfing for about 1hr plus now with 23rd Dec build and !Netsurf r13161 and not had any issues.

I will however say that I have had the occasional crash with !Netsurf with various ROM builds so I don’t think there is anything specific to the ROM build though you never know.

I would also try an older clean !Boot sequence than that on the ROOL site currently as it breaks !Boot since the live build elements have been introduced.

Doug

Since the Iyonix and BB behave identically w.r.t. the Zip drive, but completely differently w.r.t. NetSurf, I can’t see how that could possibly be true.

Ah! I didn’t realise that the Iyonix had the same issue with the Zip drive.
So that must be an entirely different issue.

If you want a new !Boot sequence, there is an older (working) build here – http://www.robheaton.co.uk/riscos/harddisc.zip

If you want a new !Boot sequence, there is an older (working) build here

Thanks, Rob.

It looks like the problem is caused by this change – https://www.riscosopen.org/viewer/revisions/logs?ident=1323987595-661104.html

I’m no expert on ObjAsm syntax, but isn’t there a syntax error in this change to the LDMRet macro in swiveneers

-        LDM$cond.FD   sp!,{$regs,pc}^
-        |
-        LDM$cond.FD   sp!,{$regs,pc}
+        LDM$cond.FD   sp!, { $regs },^   ;<== erroneous comma before ^
+        |             
+        LDM$cond.FD   sp!, { $regs }

Although that seems to be only for {CONFIG}=26, so presumably not the problem you’re all seeing? If it is wrong, I would have thought the assembler would have refused to assemble it anyway…

I’m no expert on ObjAsm syntax, but isn’t there a syntax error in this change to the LDMRet macro in swiveneers

The author of the changes has sent me a fix to test, so far, so good! He hasn’t picked up on that erroneous comma though! I’ll send him a link to your post.

Nice work WPB!

While I’m at it, is the removal of the ALIGN in poduleirq after the “AREA” statement at line 44 safe? (See diff here)

It seems to have been pruned out along with the superfluous (due to doubling up with ASMUtils) user_mode_donothing() function, but could it not leave the subsequent labels “splhi” etc. pointing a few unaligned bytes before the first instruction of the routine? Like I said, I’m no expert, so happy to be proved wrong!

Turns out the bug was in AsmUtils.
A fix has now been checked into the CVS.

I’ve built a set of ROMS that include the fix, they are available here – http://www.robheaton.co.uk/riscos/roms

Rob

Tried the 24th Dec OMAP3 ROM and !Grapevine, !Netsurf and Hermes all work OK

Great work

Doug

If you want a new !Boot sequence, there is an older (working) build here

Has anyone tried to use a new !Boot sequence recently?

If you’ve got a couple of days to spare, try copying your existing !Boot out of sight and substituting a fresh standard one.

Networking no longer works. When I click “Set” on the “Host names” page, I get an error box titled “Message from Internet Setup” with the rather bald message “Not found”. OK, give me a clue, please, RISC OS: what were you looking for?

The good news is that, after I replaced my !Boot with a standard one but before I rebooted, NetSurf was happy to browse the Lolcats site without crashing, so the bad problem I had has somehow been cured. But now the challenge is to get the thing working “properly” again, i.e. as I want it to be configured. More good news is that I was able to take a complete backup of the BeagleBoard’s hard drive to NAS before I replaced !Boot.

Now if only I can get the BB on the LAN again to access that backup… or anything at all from any other machine on the LAN… or anything at all on the Internet…

You need to install the driver for the network card.
Download EtherUSB from here – http://effarig.co.uk/riscos/

You need to install the driver for the network card.

Yup – got that. Though I wish it were included in the !Boot sequence – or, failing that, I wish the instructions for use came with it… in use for an ordinary user, it needs an Obey file, or possibly two. There are instructions somewhere in these fora, and I had followed them. Still “Not found” and the settings were not saved.

Currently, when I click on Configuration→Network→Interfaces, the “Message from Internet Setup” is “No network interfaces were detected”. I’m sure there should be an Obey file somewhere to cause the EtherUSB module to be initialised, although *EJInfo does show some stats and the EtherUSB module (0.12) is running.

I think something goes wrong earlier in the boot-up, although I don’t see any error message to say so. LanMan98 is in Apps, but issuing *Connect98 results in “File ‘connect98’ not found”. Also I see that my changes to the monitor type and settings have been forgotten on a reboot. When I try to set them again, I get “Message from Screen setup”: “Can’t save screen mode because resolution or colours is unknown”, despite having carefully configured both of them.

It looks like one or more paths is/are undefined, so configuration information can’t be saved.

If we’re going to recommend people to replace their !Boot sequence, we’re going to have to come up with one that doesn’t wreck quite so much, and doesn’t need so much experienced help to get them over the problems it creates.

There’s no Boot:Choices.PreDesk or anything like that. Alarm comes up using the default settings, ignoring my saved choices. I’ve added a PreDesk directory, so at least I can save Alarm’s choices and not get an error message instead.

Sorry, I’ll withdraw part of that: Alarm does use the saved choices.

The network settings are now being retained, and there is no “Not found” message on trying to save them. (The restart on a change of settings is very dodgy, but that’s an advanced topic for another day – and there is an easy work-round – power it down!)

I think the !Boot sequence must have a PreDesk directory added to it.

Next challenge is how to get the monitor settings to stick, and get rid of the “Can’t save screen mode because resolution or colours is unknown” message.