Thomas Olsson's MkSprite and SFMerge now available

After looking around for sources, Thomas Olsson has kindly allowed me to publish the sources for his mksprite and sfmerge, together with 32bit versions of these tools. This code had been lost, but Thomas found that a copy had been sent to Will Blatchley, who generously found the old code and made it available!

The source can be found at:

https://github.com/gerph/thomasolsson-mksprite

Releases can be found on the Releases page at:

https://github.com/gerph/thomasolsson-mksprite/releases

Thomas’s website with his other software can be found here:

http://www.armpit.dk/

I’m getting the following errors when running the ‘test’ file

Not enough memory to change to this screen mode

Postmortem requested
  Arg4: 0x00000024 36 -> [0x52454645 0x45434e45 0x21212121 00000000]
  Arg3: 0x00000024 36 -> [0x52454645 0x45434e45 0x21212121 00000000]
  Arg2: 0x0000d0a8 53416 -> [0x00000023 0x00dddddd 00000000 0x0000000d]
  Arg1: 0x0000ffb8 65464 -> [0x00001480 0x00000001 0x00000010 0x0000147c]
fc1891f0 in function tidy_sprite
  Arg2: 0x0000d0a8 53416 -> [0x00000023 0x00dddddd 00000000 0x0000000d]
  Arg1: 0x000117f8 71672 -> [0x00001a20 00000000 0x00000010 0x00000010]
9d28 in function post_process
  Arg2: 0x0000fa48 64072 -> [0x0000fa74 0x0000fa7d 0x0000fa80 0x0000fa82]
  Arg1: 0x00000008 8 -> [0xfdfdfdfd 0xfdfdfdfd 0x21212121 0x4c4c554e]
a6b8 in function main
  Arg2: 0x0000a1ec 41452 -> [0xe1a0c00d 0xe92dd9f3 0xe24cb004 0xe24dcf6e]
  Arg1: 0x0000f540 62784 -> [0x70736b6d 0x65746972 0x206d2d20 0x732d2034]
fc1790a0 in shared library function
a878 in anonymous function

Running on Raspberry Pi and also tried on RPCEmu. Anyone know what’s wrong?

thanks

How are you doing it? If via an Obey file, maybe it needs a larger WimpSlot?

OK so there were no instructions with MkSprite so I had to rub the crystal ball and came up with the following:

In the download MkSprite-1/10/zip there is (among other files) a directory called Library. I copied the two files MkSprite and SFMerge to !Boot.Library.

Also inside MkSprite-1/10/zip there is a directory called Examples, and inside that is an obeyfile called test. When that is double clicked it gives the errors above.

Dragging the ‘Next’ slider in the Tasks window to the maximum it can go (here it’s 28640K) doesn’t cure it, is that what you mean by setting a larger WimpSlot or something else?

thanks

A common fault with old code is mistaking too much memory for too little memory.

@ Charles

Thanks, I have tried this and I can confirm there seems to be some bug when running it on RO 5 (I haven’t tested it yet on 26bit RO releases).

Also, when trying to build the sources there are a number of problems:

1) Your sources comes with the MakeFile (MakefileOriginal) still with Dynamic Dependecies from yoru own builds, these should be cleaned up and test the build from complete clean in order to avoid Norcroft to play funny tricks. These are causing the build to fail on RO 5 (can’t find C:h.SWINames )

2) I have cleaned them manually and the build still fails, this time with a bunch of warnings, plus quite a few errors, too many to reports here, but we can swithc to emails to list them all.

Hope this helps,

Ta for the report. Please remember that reporting issues in the GitHub issues systems is far more reliable than anything in these forums. For a start, it reports the issues in the place where they are relevent, and – more usefully – it actually notifies you when there’s an issue.

The sources is pretty much as supplied by Thomas, but made to work as a 32bit build. Warnings are as supplied. There should be no errors. The original makefiles are, again, as supplied. If they need stripping, feel free to raise a PR for that (Future note: I’ve done this now). I don’t use the original makefiles, as they were dependant on things that don’t exist, and are provided purely for curiousity.

The report that came by email looks like you’ve not built the library that it uses, and that’s explictly what it’s telling you by saying that it cannot find tholib.h. Build the library and reference it in your makefile and it should build fine.

I’ll have a look at the builds – it’s always possible that I’ve broken something in rebuilding them.

I can confirm that it doesn’t work in RISC OS 5 using the docker RPCEmu I use. The tidy_sprite calls aren’t that complicated, so it’s surprising.

Given that we never change mode, the error ‘Not enough memory to change to this screen mode’ seems wrong. Might mean that it thinks the sprite area is too small. Investigating anyhow.

The code creates a sprite using a mode selector with size 36×36 using 16 colours, with the mode flags set to NColours = 128 (this is odd), XEigFactor = 1 and YEigFactor = 1.

If the NColours value is reset to 15 (which is what it should be for a 16 colour mode), it works. Presumably this has been tightened up in RISC OS 5, but the error message hasn’t been made clear that the mode selector parameters were invalid, I guess.

Please remember that reporting issues in the GitHub issues systems is far more reliable than anything in these forums.

I agree, thought you’d prefer on here, I too prefer GitHub, so will move every further chats to open issues on GitHub, thanks! :)

Code now fixed, and version 1.11 released.

You can find updated versions available here: https://github.com/gerph/thomasolsson-mksprite/releases/tag/v1.11