Odd message, you can try to see if they call int15 and require some
specific return code for some function (using the debugger-enabled
version of dosbox, use "bpint 15 *" to catch it).
Or maybe they look at the seg/offset of the int15 handler, if you are
lucky you can check this by a "bpint 21 35" (that should be the get
int vector).

Thanks for the help wd, but I'm afraid I do not have the skills to find what you are looking for. I am an electronics technician, not a programmer so the only experience I have with debugging is with very small programs(30 lines and less) I wrote for the 8086 CPU(and x51 MCU, but that's not relevant). I don't know if it's because the 80286(and up) are different but there are a lot of instructions I don't understand, and everything looks so different from what I'm used to (in the 8086 the 4 registers AX,BX,CX and DX are all 16bit, and here I see they are 32bit... and more).
anyway, I attached two pictures, one is the last int 15 call before the error appears, and the other is the last function 35 from the 21h software interrupt before the error appears.
I don't know if the few instruction visible in the pictures will help you see what's wrong, but here it is anyway(or maybe you could tell me what exactly to look for?).

http://www.delorie.com/djgpp/doc/rbinter/ix/15/83.html
bios wait event interval.

I recently added some more calls to that function although it appears to be using the call 0 which didn't change.
Try this build of dosbox
http://pcnwstage.phys.rug.nl/dosbox-shift.exe.bz2
and please report back if it changed anything.

I assume you traced into the int15, at least it looks like that in the
screenshot. If so, they seem to have revectored int15 already to
0x12cd:0x0f27 (the retf2 replaces an iret more or less).

Maybe you can pm the link to me or Qbix.

Cool, now it works! 😁
Thanks Qbix!

BTW, is this already in the CVS? or will it be added?

Thanks again
Og

It's allready in the cvs.
Thank you for reporting back.

So what is the percise problem which the CVS fixes?

for int15: cancel wait interval should always succeed