bio printf: Avoid using rounding errors in range check

There is a problem casting ULONG_MAX to double which clang-10 is warning about.
ULONG_MAX typically cannot be exactly represented as a double.  ULONG_MAX + 1
can be and this fix uses the latter, however since ULONG_MAX cannot be
represented exactly as a double number we subtract 65535 from this number,
and the result has at most 48 leading one bits, and can therefore be
represented as a double integer without rounding error.  By adding
65536.0 to this number we achive the correct result, which should avoid the
warning.

The addresses a symptom of the underlying problem: we print doubles via an
unsigned long integer.  Doubles have a far greater range and should be printed
better.

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/11955)

2 files changed, 41 insertions, 1 deletions

diff --git a/crypto/bio/b_print.c b/crypto/bio/b_print.c
index 0d6fafcc2d..6b995f8233 100644
--- a/crypto/bio/b_print.c
+++ b/crypto/bio/b_print.c
@@ -635,7 +635,13 @@ fmtfp(char **sbuffer,
             fvalue = tmpvalue;
     }
     ufvalue = abs_val(fvalue);
-    if (ufvalue > ULONG_MAX) {
+    /*
+     * By subtracting 65535 (2^16-1) we cancel the low order 15 bits
+     * of ULONG_MAX to avoid using imprecise floating point values.
+     * The second condition is necessary to catch NaN values.
+     */
+    if (ufvalue >= (double)(ULONG_MAX - 65535) + 65536.0
+            || !(ufvalue == ufvalue) /* NaN */) {
         /* Number too big */
         return 0;
     }
diff --git a/test/bioprinttest.c b/test/bioprinttest.c
index 14f0bfe52d..3dd5b3efa2 100644
--- a/test/bioprinttest.c
+++ b/test/bioprinttest.c
@@ -241,14 +241,48 @@ static int test_fp(int i)
     return r;
 }
 
+extern double zero_value;
+double zero_value = 0.0;
+
 static int test_big(void)
 {
     char buf[80];
+    double d, z, inf, nan;
 
     /* Test excessively big number. Should fail */
     if (!TEST_int_eq(BIO_snprintf(buf, sizeof(buf),
                                   "%f\n", 2 * (double)ULONG_MAX), -1))
         return 0;
+
+    d = 1.0;
+    z = zero_value;
+    inf = d / z;
+    nan = z / z;
+
+    /*
+     * Test +/-inf, nan. Should fail.
+     * Test +/-1.0, +/-0.0. Should work.
+     */
+    if (!TEST_int_eq(BIO_snprintf(buf, sizeof(buf),
+                                  "%f", inf), -1)
+            || !TEST_int_eq(BIO_snprintf(buf, sizeof(buf),
+                                         "%f", -inf), -1)
+            || !TEST_int_eq(BIO_snprintf(buf, sizeof(buf),
+                                         "%f", nan), -1)
+            || !TEST_int_eq(BIO_snprintf(buf, sizeof(buf),
+                                         "%f", d), 8)
+            || !TEST_str_eq(buf, "1.000000")
+            || !TEST_int_eq(BIO_snprintf(buf, sizeof(buf),
+                                         "%f", z), 8)
+            || !TEST_str_eq(buf, "0.000000")
+            || !TEST_int_eq(BIO_snprintf(buf, sizeof(buf),
+                                         "%f", -d), 9)
+            || !TEST_str_eq(buf, "-1.000000")
+            || !TEST_int_eq(BIO_snprintf(buf, sizeof(buf),
+                                         "%f", -z), 8)
+            || !TEST_str_eq(buf, "0.000000"))
+        return 0;
+
     return 1;
 }