VAX: one more VAX fp doc tweakage

Frankly quite amazing how many times one can read one's
own writing and miss things.

diff --git a/perl.h b/perl.h
index 4a6c4f1..dddf97d 100644 (file)
--- a/perl.h
+++ b/perl.h
@@ -6778,13 +6778,15 @@ extern void moncontrol(int);
  *
  * The ordering of the parts in VAX floats is quite vexing.
  * In the below the fraction_n are the mantissa bits.
+ *
  * The fraction_1 is the most significant (numbering as by DEC/Digital),
  * while the rightmost bit in each fraction is the least significant:
  * in other words, big-endian bit order within the fractions.
+ *
  * The fraction segments themselves would be big-endianly, except that
- * within 32 bit segments the less significant comes first, the more
+ * within 32 bit segments the less significant half comes first, the more
  * significant after, except that in the format H (used for long doubles)
- * the first fraction segment is on its own.  As promised, vexing.
+ * the first fraction segment is alone, because the exponent is wider.
  * This means for example that both the most and the least significant
  * bits can be in the middle of the floats, not at either end.
  *
@@ -6794,10 +6796,10 @@ extern void moncontrol(int);
  * http://h71000.www7.hp.com/doc/82final/6443/6443pro_028.html
  * (somebody at HP should be fired for the URLs)
  *
- * F   fraction_2:16 sign:1 exp:7  fraction_1:7
+ * F   fraction_2:16 sign:1 exp:8  fraction_1:7
  *     (exponent bias 128)
  *
- * D   fraction_2:16 sign:1 exp:7  fraction_1:7
+ * D   fraction_2:16 sign:1 exp:8  fraction_1:7
  *     fraction_4:16               fraction_3:16
  *     (exponent bias 128)
  *