IIRC the checksums are not particularly strong and it should be possible to develop an add-in that by wasting some space, brute-forces the checksum so that it matches with conv.g3a in everything except size (unless size is checked too?). There would not be much point in doing it, except for using the space conv.g3a takes in something more useful. And it would be always possible for Casio to "fix" the problem in a later OS version, annoying them a bit and rendering the whole effort useless.

I totally agree with you on the above.

Regarding the size - it is not checked: any valid add-in is accepted as long as the previously mentioned few requirements are met. I have not managed to create a not valid g3a to be accepted by runmat so not sure about those - but one of my attempts was to create an addin which does not get accepted to show up in the main menu but get's accepted by runmat. Maybe it's possible, someone let me know if they succeed.

Regarding brute forcing - I agree the checksum32 is very weak but annoyingly I see no way of reducing the checksum. So the problem starts once you have an addin with any larger checksum than conv.g3a. Or am I wrong about it or do you mean making a really big addin and then checksum gets maxed out at some point and restart from the beginning - I think this would not work as to loop the checksum about 16mb of data/file size would need to be added which would hit the storage limit on the device. Let me know if I misunderstood you.

While revisiting this post I actually had a nice idea of very neatly and quickly integrating runmat and your eigenmath - so would take you very little time to achieve this and I would find time to assist on this if you are happy to revisit your eigenmath project at all. Let me know please and I either PM you or post on your eigenmath thread.

Many thanks for everything again.