@nm I mentioned multiple times that I didn't claim that there would be any benefit to doing this, and that I only wrote the code so that you could safely explore what would happen if you stopped using remember tables. The removal of options remember or cache from approximately 1000 Library procedures is the only thing that my code does.

@vv From an asymptotic expansion with leading term sqrt(3/n), how do you get a limit of 1 instead of 0? Indeed, direct numeric computation of (sin@@n)(1.) for large n suggests very slow convergence to 0. Also, sin(x) < x for any x > 0.

@acer Thanks. It seems that your experience of it is quite similiar to mine. A little testing in Maple 2025 shows that evalf undertands it, but allvalues does not. 

I think that any usage of it by commands that generate RootOfs should be reported.

I assume that some other command returned to you that RootOf expression. What command was that? The indexing with repect to real roots is I believe very little known, and I don't think that other commands understand it when that RootOf is passed on to them.

If y is a nonnegative real, then RootOf(_Z^2 - y, index= 1) denotes the positive square root of y and index= 2 denotes its negative (positives come before negatives in the index order).

@acer The set Sol contains those inequalities, and the solve call is 

solve(Sol union area)

By the way, can you find any documentation on the real argument to RootOf? I saw some many years ago. It was quite sketchy, and now I can't find it at all.

@Scot Gould What makes you think "The folks of Maplesoft look at questions in MaplePrimes when making changes"? This Question about a dark theme has been asked at least twice in the past few years. 

As my vision is declining with age, I desperately need a dark background. It should be common sense that light letters on a dark background are far easier to read than dark letters on a light background because in the latter case the vast majority of the light hitting your retinae is superfluous information.

@Scot Gould @dharr @nm  Windows 11 has quick snaps to any corner-quarter screen also. 

I confirm that this is true in Maple 2025 Windows 11 also. No problem with older Maple.

Please upload the worksheet that has this problem using the green uparrow on the editor's toolbar. 

@janhardo I don't see any connection between your Reply and the OP's Question or its attached worksheet. Perhaps you were intending to answer some other Question on this forum.

You simply need to learn how to do better searches with Google (or other search engine). The following searches using Google in Google Chrome returned exactly the info that you wanted as the the top hits. These were my 1st attempts: no guessing, no refining.

largest maplesoft product

earliest maple math software release

Changing the name won't help, because all the old web pages will still contain "Maple", which'll thus be the word indexed by Google.

@sursumCorda Your map2(`?()`, [f], L) is a great idea. It can be further shortened to

[`?()`]~(f, L)

In addition to @mmcdara 's fine Answer, I want to caution you to be careful not to confuse these two error messages:

1. initial Newton iteration is not converging
2. Newton iteration is not converging

The 2nd is much more difficult to fix than the 1st. Fixing the 1st often leads to the 2nd; nonetheless, you should still fix the 1st before attempting to fix the 2nd. 

@aroche I believe that the following much simpler is evaluation shows a more-proximal cause of the disappointing FAIL that you obtained for the cos/sin problem:

is(Or((x,-x)=~ abs(x))) assuming x::real;  #1
                              FAIL

Compare with this correct evaluation:

is(signum(x) in {-1,0,1}) assuming x::real;   #2
                              true

It should be easy to convert problem #1 to problem #2, and I think that this alone may be enough to solve the cos/sin problem and numerous others involving a disjunction of the branches of sqrt under a real assumption.

The above Answer was written before I saw that simulated rolls of a standard 6-sided die were a requirement for you. The above code can be easily modified to accomodate that. A roll d in 1..6 can be converted to your {-1, 1} by the simple function sign(d-4), so the subs is still not needed. (See help page ?sign, and note that sign(0)=1, and don't confuse sign with signum.)

N:= 1000:
randomize(); #optional 
D6:= rand(1..6): #6-sided die simulator
Y:= [seq['scan'= `+`]](sign~([seq['scan'= D6]](1..N) -~ 4)):
plots:-listplot(Y, size= 200*[4,1], thickness= 0.7, color= red);