ChangeFSI hang

If you try and process this using ChangeFSI you get a crash.
What you have to do is select jpg output, save choices, then select sprite output and save choices then try and process it.
OR run double click and let it run for the first time.

The sure way for me is boot up and just double click it, ChangeFSI pans (with sprite output saved)

More info coming once my machine comes back….

Modinfo is reporting this as the Utility Module, but I’m not sure as that ran after the reboot as attempting to do anything afterwards causes a reset :-(

I’ve just had an “Internal error abort on data transfer at &000874D4 (internal number 9082)” – that’s a new one!
Then If I try again I get “Internal error abort on data transfer at &FC041584”
Then it’s really bad news, I get a reset of the machine from what I can tell.

*modules
No. Position Workspace Name
1 FC01F44C 00000000 UtilityModule
2 FC041C3C 20000014 PCI
3 FC043B44 200001D4 FileSwitch
4 FC050BB0 20001694 ResourceFS

But also I’m getting a lot of these errors when using NewsUK (the ChangeFSI handling)
Attempt to use badly nested error handler (or corrupt R13)
Division by zero (internal number 9073)

Also it seems to be hanging processing jpg’s a lot more now, if I alt-break it says “Press Stop to terminate Unknown” which is the ChangeFSI process I’ve spawned.

Latest ROM and Boot image for a pandaboard ES

If you try and process this using ChangeFSI you get a crash.

The hair looks a little too red (even accounting for background shading) to be true to Kusanagi. Still, I like the fact that she’s dressed (!) and not carrying a gun as big as her torso (!). The city-scape reflection in the goggles doesn’t quite match the windowed room appearance, but the lightwork is quite pleasing. :-)

Does ChangeFSI bork on this version? http://i.imgur.com/ttJhfay.jpg

Oh… Have you tried giving ChangeFSI a really big WimpSlot?

If you try and process this using ChangeFSI you get a crash.

With a feeling of deja vu, try Set ChangeFSI$Cache 16384K

Works here on Raspberry Pi, OS5.21 (20-Jun-14).

Rick: It doesn’t crash with that version.
It’s not my favourite backdrop of the Major by a long way, but I got bored with my last one.

I’ll do more testing, with wimpslot etc. later…

David: Yeah it is a feeling of deja vu, sorry.
Looks like this is what is breaking newsuk occasionally too, so I’ll alter the parameters I’m passing newsuk.
Would it be possible to change this to be the default on changefsi as the results can be fairly catastrophic.

I assume this is the same as -max when using the commandline?
Looks like it’s not, I’m still getting the same issue with NewsUK although the complete crash doesn’t happen just the internal error

If I was passing -nomode, -lock and -max it still hung.
I’ve changed it to only pass -nomode and it seems to behave much better (as I have to pass that)
I still get an occasional hang in ChangeFSI, but I’m suspecting there is a duff download sometimes, which gets passed to ChangeFSI which doesn’t like it.
I’ll fix that in v2.
Looks like I’ll have to set the ChangeFSI Cache value in my !Run file for stability too for the moment.

It’s really weird how it crashes on one and then not on another, the size in NewsUK jpgs etc are small as well.

Looks like there will be another newsuk update :-(
dammit!

David, you may want to check the !ReadMe on the NewsUK update when it comes ;-)

Unfortunately it’s not fixed the problem, it seems to fix the problem for large files – but the underlying issue whatever it is, is still lurking away. There seems to be some kinds of jpgs that cause this issue as I’m still getting the same issue (no serious crash after) where I get “Internal error abort on data transfer at &000874D4 (internal number 9082)”

I suspect that you are running a recent RISC OS, with Basic at least v1.57? If so, the ‘internal number 9082’ should be the Basic Line number where the error happened – which would have been given (wrongly) as the last line in the program before v1.57.

If you give the version of ChangeFSI you are running, then the line number may be a clue about the cause of the problem.

I’m running nightly updates of RISC OS on a pandaboard…
So basic is v1.59 and ChangeFSI is version 1.47
Ah ha! The problem is with the 32bpp rendering. If you select 32bpp rendering in newsuk it barfs a lot…..Corruption of stack or internal error

That’s passing S32 to changeFSI

Need a cat …is this one planning world domination or just grumpy?

When I get round to it I’ll disable 32bpp as it does look like it’s broken, and there isn’t much point of 32bpp for the graphics in newsuk.