jkesselm 02/02/05 07:11:26
Modified: java/src/org/apache/xpath/objects XStringForFSB.java
Log:
Bug 5346: One more (last, I hope) edge case.
Revision Changes Path
1.13 +6 -5
xml-xalan/java/src/org/apache/xpath/objects/XStringForFSB.java
Index: XStringForFSB.java
===================================================================
RCS file:
/home/cvs/xml-xalan/java/src/org/apache/xpath/objects/XStringForFSB.java,v
retrieving revision 1.12
retrieving revision 1.13
diff -u -r1.12 -r1.13
--- XStringForFSB.java 5 Feb 2002 14:19:55 -0000 1.12
+++ XStringForFSB.java 5 Feb 2002 15:11:25 -0000 1.13
@@ -1050,12 +1050,13 @@
// We have a potential overflow issue that we need
// to guard against. See Bugzilla 5346.
//
- // %REVIEW% MUST BE RECONSIDERED since this may not be
- // fully compliant with the IEEE floating-point spec. (On the
- // other hand, it's wicked fast compared to the usual solution,
- // and works fine for all but a very few edge cases.)
+ // %REVIEW% MUST BE RECONSIDERED. I _think_ the truncation of
+ // the 64-bit long before it overflows is well within the acceptable
+ // error range of the double's 53-bit mantissa...
if(longResult>MAX_NO_OVERFLOW_RISK)
- ++overflow;
+ {
+ if(digitType==0) ++overflow;
+ }
else
{
longResult=longResult*10 + (c - '0');
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
