make BigFloat(x::BigFloat) respect global precision (fix #20766)

NEWS.md

@@ -9,6 +9,9 @@ New language features
 Language changes
 ----------------
 
+ * the constructor `BigFloat(::BigFloat)` now respects the global precision setting and always
+ returns a `BigFloat` with precision equal to `precision(BigFloat)` ([#29127]).
+
 
 Standard Library Changes
 ------------------------

base/mpfr.jl

@@ -116,7 +116,17 @@ BigFloat(x)
 widen(::Type{Float64}) = BigFloat
 widen(::Type{BigFloat}) = BigFloat
 
-BigFloat(x::BigFloat) = x
+function BigFloat(x::BigFloat)
+ if precision(BigFloat) == precision(x)
+ x
+ else
+ z = BigFloat()
+ ccall((:mpfr_set, :libmpfr), Int32, (Ref{BigFloat}, Ref{BigFloat}, Int32),
+ z, x, ROUNDING_MODE[])
+ z
+ end
+end
+
 
 # convert to BigFloat
 for (fJ, fC) in ((:si,:Clong), (:ui,:Culong))

test/mpfr.jl

@@ -35,6 +35,20 @@ import Base.MPFR
 
  @test typeof(BigFloat(1//1)) == BigFloat
  @test typeof(BigFloat(one(Rational{BigInt}))) == BigFloat
+
+ # BigFloat constructor respects global precision when not specified
+ let prec = precision(BigFloat) < 16 ? 256 : precision(BigFloat) ÷ 2
+ xs = Real[T(1) for T in (Int, BigInt, Float32, Float64, BigFloat)]
+ f = xs[end]
+ @test BigFloat(f) === f # no-op when precision of operand is the same as global's one
+ setprecision(prec) do
+ @test precision(xs[end]) != prec
+ for x in xs
+ @test precision(BigFloat(x)) == prec
+ @test precision(BigFloat(x, prec ÷ 2)) == prec ÷ 2
+ end
+ end
+ end
 end
 @testset "basic arithmetic" begin
  tol = 1e-12