After upgraded to 1.9.4. Diff the gnumeric file with the previous
1.8.3
- <gnm:Cell Row="77" Col="5" ValueType="40" ValueFormat="0.00_
">5925.57</gnm:Cell>
+ <gnm:Cell Row="77" Col="5" ValueType="40" ValueFormat="0.00_
">5925.5699999999997</gnm:Cell>
@@ -4456,7 +4457,7 @@
- <gnm:Cell Row="82" Col="5" ValueType="40" ValueFormat="0.00_
">1408.4</gnm:Cell>
+ <gnm:Cell Row="82" Col="5" ValueType="40" ValueFormat="0.00_
">1408.4000000000001</gnm:Cell>
@@ -4514,7 +4515,7 @@
- <gnm:Cell Row="95" Col="5" ValueType="40" ValueFormat="0.00_
">1208.63</gnm:Cell>
+ <gnm:Cell Row="95" Col="5" ValueType="40" ValueFormat="0.00_
">1208.6300000000001</gnm:Cell>
I did not compile with long double support. It looked like a regression
because the actual ieee bit pattern for 5925.57 and 5925.5699999999997 are
exact the same yet the ver 1.8.3 use minimum number of decimal digit to
represent the same ieee double. There is an algorithm to find the decimal
number with minimum number of digit for the same bit pattern in ieee
representation. Or the previous version in fact did not strive for minimum
digits representation but just looked like that way?
--
regards,
====================================================
GPG key 1024D/4434BAB3 2008-08-24
gpg --keyserver subkeys.pgp.net --recv-keys 4434BAB3
唐詩282 白居易 後宮詞
淚濕羅巾夢不成 夜深前殿按歌聲 紅顏未老恩先斷 斜倚薰籠坐到明
_______________________________________________
gnumeric-list mailing list
gnumeric-list@gnome.org
http://mail.gnome.org/mailman/listinfo/gnumeric-list
