http://issues.apache.org/SpamAssassin/show_bug.cgi?id=4742
------- Additional Comments From [EMAIL PROTECTED] 2005-12-19 18:03 ------- > Unfortunately, thats the easy case, here is what happens in reality: > If your threshold is 5, then 5.1 is spam and 5.0 is ham, as well as 4.9 is ham. No. What happens in *reality* is that an indefinite-precision number is compared to a definite-precision constant, in your case 5.00000000000. Any number in internal storage format with a value less than the 5.000000000000000 will be shown as ham, for instance 4.999999999999999. Any value of 5. or greater is spam. Contrary to your statement, 5.000000000000 is SPAM, not HAM. *AFTER* this comparison is made, using the internal values, the *internal* value is rounded to a *display* value, using the *number of rounding digits you specify*. The number of display rounding digits you specify does not affect the comparison; nor should it. It is a DISPLAY option, not a COMPARISON option. If you wish to set an integral display rounding, and still want to know if the mail is ham or spam, there are three possible ways: 1. Examine the X-Spam_Status Yes/No header. It represents the results of the internal comparison and does not lie. 2. Use an integral spam threshold (not fractional) and examine the number of stars. The number of stars do not lie IF the spam threshold is integral. 3. Examine the Subject header for the spam flag if configured, typically *** SPAM ***. It also does not lie. If you want an *erroneous* determination of whether the message is ham or spam, set the rounding to less than three fractional digits and examine the spam score. Since it will round up or down to match the requested number of fractional digits, it can lie any time the *actual* score is close to the threshold, and gets rounded to the other side due to loss of siginicant digits in the rounding. ------- You are receiving this mail because: ------- You are the assignee for the bug, or are watching the assignee.
