Pi4 print to PDF problem

Setting up my new Pi4. When it comes to printing to PDF, everything goes fine until I reach the Create PDF dbox. On dragging the PDF file icon to a directory, the Pi4 hangs completely; I can’t break out of it. After a very long delay, an error message comes up from PrintPDF Child: ‘PrintPDF Child may have gone wrong’ (not very informative). What needs fixing? Any help much appreciated.

Running PrintPDF 1.12, Printers 1.89, Ghostscript 8.71. The file printed was from OvationPro, which worked fine on my late lamented Pi3.

Maybe a newer Ghostscript helps…
https://www.riscosports.co.uk/downloads.html.

After a very long delay, an error message comes up from PrintPDF Child: ‘PrintPDF Child may have gone wrong’ (not very informative).

No, but you could easily make it so. That’s a standard RISC OS error which is hiding the error raised by PrintPDF Child¹, and there should be a “Describe”, “More Info”, “Details” (or something) button to click. Do so, and tell us what the real error is.

Maybe a newer Ghostscript helps…

Maybe, but I doubt it.

I’m going to guess that when we finally get the real error message, it’s a segmentation fault (or whatever the RISC OS equivalent is). PrintPDF is set up for Martin W’s GhostScript ports, and Chris G’s require a lot more memory.

My suggestion would be to start by increasing the memory allocation in PrintPDF’s choices to a larger figure. I’ve no idea off the top of my head what the value should be for the latest versions of GhostScript – can anyone else advise?

(I’m not near a RISC OS machine at the moment, in case it’s not obvious.)

¹ Because when they released the RiscPC, someone at Acorn thought that showing users real error messages would be off-putting, apparently.

| tell us what the real error is

Thanks for the prompt, but the message displayed on clicking Describe is still incomprehensible (to me, anyway):

Internal error: undefined instruction at &000100F0

Regarding memory allocation, I have already doubled the default, to 16384K. The file attempted to be printed was a 270K OvPro document, containing some pictures but no unusual fonts.

Internal error: undefined instruction at &000100F0

Ah. In which case, try Raik’s suggestion and give 9.27 a whirl: either you’ve got an ARMv7 version of GhostScript on the ARMv8 Pi 4, or there’s something else a lot more complex going on.

ETA: While it may be obscure, “undefined instruction” is a lot more helpful to those helping you than “application may have gone wrong…”. Originally, all I could do was guess on probabilities as the latter message is meaningless (thanks, Acorn) – and lack of memory is the usual gotcha by far. With the real error message, however, it’s much more likely to be a rogue SWP instruction on a machine which doesn’t have them.

I just printed a 15 page document OK as PDF on disc from the now-free Ovation Pro 2.78g on my Raspberry Pi 3 / RISC OS 5.29 using a Chris G GhostScript (which transiently took 29MB). There’s a new GhostScript9 package in PackMan that uses VFP too (July 2021 build; 9.27 – though its !Boot still sets up variables proclaiming the old version).

Have you upgraded the Ovation Pro, Richard? It might have some instructions that aren’t compatible with the ARMv8 Pi 4.

[Edit: I’d also make sure the SharedUnixLibrary is up-to-date – check PackMan for updates.]

Increased the memory to about 130Mb; Message from PrintPDF Child now described as: SWI &204353 not known

I’ll try 9.26.

SWI &204353 not known

If I’m not mistaken, that’s an invalid number (I believe they’re 20-bit, which means they can’t exceed five digits). Can you check that you haven’t introduced an extra digit when copying the number down?

The SWI instruction has 24 bits, but ISTR all RISC OS SWIs used to have bits [23..20] clear. The SWI number passed through OS_CallASWI and its friends could be any 32 bit value, but the kernel’s handler does mask off the top 8 bits.

SWI &204353 not known

Blind stab in the dark (with ears covered):
Is that a faulty attempt to add &20000 to produce the X form SWI?

It looks more to me like text.

“ CS”

“SC " – middle of of RISC OS? Probably a SWI being called using CallASWI, loaded the register from the wrong place.