diff --git a/base/strings/unicode.jl b/base/strings/unicode.jl
index 5d59b0fc3bff4..5dcc41b030473 100644
--- a/base/strings/unicode.jl
+++ b/base/strings/unicode.jl
@@ -6,7 +6,7 @@ module Unicode
 import Base: show, ==, hash, string, Symbol, isless, length, eltype,
              convert, isvalid, ismalformed, isoverlong, iterate,
              AnnotatedString, AnnotatedChar, annotated_chartransform,
-             @assume_effects, annotations
+             @assume_effects, annotations, is_overlong_enc
 
 # whether codepoints are valid Unicode scalar values, i.e. 0-0xd7ff, 0xe000-0x10ffff
 
@@ -262,17 +262,15 @@ julia> textwidth('⛵')
 2
 ```
 """
-function textwidth(c::AbstractChar)
-    ismalformed(c) && return 1
-    i = codepoint(c)
-    i < 0x7f && return Int(i >= 0x20) # ASCII fast path
-    Int(ccall(:utf8proc_charwidth, Cint, (UInt32,), i))
-end
+textwidth(c::AbstractChar) = textwidth(Char(c)::Char)
 
 function textwidth(c::Char)
-    b = bswap(reinterpret(UInt32, c)) # from isascii(c)
+    u = reinterpret(UInt32, c)
+    b = bswap(u) # from isascii(c)
     b < 0x7f && return Int(b >= 0x20) # ASCII fast path
-    ismalformed(c) && return 1
+    # We can't know a priori how terminals will render invalid UTF8 chars,
+    # so we conservatively decide a width of 1.
+    (ismalformed(c) || is_overlong_enc(u)) && return 1
     Int(ccall(:utf8proc_charwidth, Cint, (UInt32,), c))
 end
 
diff --git a/test/strings/util.jl b/test/strings/util.jl
index bb87881bbaa1d..9ced27ee3f8d0 100644
--- a/test/strings/util.jl
+++ b/test/strings/util.jl
@@ -8,6 +8,8 @@ SubStr(s) = SubString("abc$(s)de", firstindex(s) + 3, lastindex(s) + 3)
         @test textwidth(c^3) == w*3
         @test w == @invoke textwidth(c::AbstractChar)
     end
+    @test textwidth('\xc0\xa0') == 1 # overlong
+    @test textwidth('\xf0\x80\x80') == 1 # malformed
     for i in 0x00:0x7f # test all ASCII chars (which have fast path)
         w = Int(ccall(:utf8proc_charwidth, Cint, (UInt32,), i))
         c = Char(i)