fix reinterpret(Char, ::UInt32) for "unnatural" values (fix #29181) (#29192)

This code was assuming that character values only have bit-patterns
that decoding a string can produce, but of course `reinterpret` can
produce any bit pattern in a `Char` whatsoever. The fix doesn't use
that assumption and only uses the cache for actual ASCII characters.

(cherry picked from commit 88f74b7)

Author: StefanKarpinski

src/datatype.c

@@ -688,8 +688,9 @@ static jl_value_t *boxed_char_cache[128];
 JL_DLLEXPORT jl_value_t *jl_box_char(uint32_t x)
 {
     jl_ptls_t ptls = jl_get_ptls_states();
-    if (0 < (int32_t)x)
-        return boxed_char_cache[x >> 24];
+    uint32_t u = bswap_32(x);
+    if (u < 128)
+        return boxed_char_cache[(uint8_t)u];
     jl_value_t *v = jl_gc_alloc(ptls, sizeof(void*), jl_char_type);
     *(uint32_t*)jl_data_ptr(v) = x;
     return v;

test/char.jl

@@ -256,3 +256,33 @@ Base.codepoint(c::ASCIIChar) = reinterpret(UInt8, c)
     @test_throws MethodError write(IOBuffer(), ASCIIChar('x'))
     @test_throws MethodError read(IOBuffer('x'), ASCIIChar)
 end
+
+@testset "ncodeunits(::Char)" begin
+    # valid encodings
+    @test ncodeunits('\0')       == 1
+    @test ncodeunits('\x1')      == 1
+    @test ncodeunits('\x7f')     == 1
+    @test ncodeunits('\u80')     == 2
+    @test ncodeunits('\uff')     == 2
+    @test ncodeunits('\u7ff')    == 2
+    @test ncodeunits('\u800')    == 3
+    @test ncodeunits('\uffff')   == 3
+    @test ncodeunits('\U10000')  == 4
+    @test ncodeunits('\U10ffff') == 4
+    # invalid encodings
+    @test ncodeunits(reinterpret(Char, 0x80_00_00_00)) == 1
+    @test ncodeunits(reinterpret(Char, 0x01_00_00_00)) == 1
+    @test ncodeunits(reinterpret(Char, 0x00_80_00_00)) == 2
+    @test ncodeunits(reinterpret(Char, 0x00_01_00_00)) == 2
+    @test ncodeunits(reinterpret(Char, 0x00_00_80_00)) == 3
+    @test ncodeunits(reinterpret(Char, 0x00_00_01_00)) == 3
+    @test ncodeunits(reinterpret(Char, 0x00_00_00_80)) == 4
+    @test ncodeunits(reinterpret(Char, 0x00_00_00_01)) == 4
+end
+
+@testset "reinterpret(Char, ::UInt32)" begin
+    for s = 0:31
+        u = one(UInt32) << s
+        @test reinterpret(UInt32, reinterpret(Char, u)) === u
+    end
+end