an attached exe/com file showing it would be handy.
4.99999999999 =5 in computer stuff. especially if it makes use of the fpu.
But maybe there is some rounding method going wrong

Hmm... I can not confirm that bug. I tried this with a QBASIC from Win95OSR2:

110 AT$ = "05"
220 AT# = VAL(AT$)
330 PRINT AT#

Both DOSBox 0.63 and WinXP returned the same result: 5.

Edit: Ahh 😊 I did not know that QuickBasic != QBasic.

9642394: FPU:ESC 6:handled group 1 subfunction 1
fpumul 300000000000000000.000000 0.100000
output= 29999999999999996.000000

this is the problem.
due to huge order difference rounding errors apear (as we use less precission as a real fpu (64 vs 80 bits).
Will see if I can do some order and rescaling before multiplication
(or put the numbers in a 80 bit value before multiplication)

the dosbox code doing the fpumul:

1a*=b;
2LOG("output=%f",a);

Up to then you could
a) use qbx which works (or qbasic)
b) use PRINT USING "#.###"; AT# or similar

Seems that QuickBasic has a different output
precision (dynamic?) than qbx and qbasic.

wd

problem is that a 80 bit 0.1 is read in as
0.0999999999....
as they multiply with a very large
30000000000000000000
it get's slightly off.
they do this operation 2 times. that gives the small difference.
I don't exaclty see at the moment why 4 gives correct answers then.

hmmm. will be tricky. Let's just hope load80 bits has a bug 😀