NetRadiant does a lot of diff noise toying with multiple representation of zero
Look at that diff:
-( -368 48 104 ) ( -368 40 24 ) ( -368 48 24 ) shared_ex/base1a 0.0 0.0 0.0 0.5 0.5 134217728 0 0
-( -240 48 104 ) ( -368 48 24 ) ( -240 48 24 ) shared_ex/wall_panel7 64.0 36.5714111328 0.0 -0.25 0.2187499851 134217728 0 0
-( -240 40 104 ) ( -240 48 24 ) ( -240 40 24 ) shared_ex/base1a 0.0 0.0 0.0 0.5 0.5 134217728 0 0
-( -368 40 104 ) ( -240 40 24 ) ( -368 40 24 ) shared_ex/base1a 160.0 0.0 0.0 -0.5 0.5 134217728 0 0
+( -368 48 104 ) ( -368 40 24 ) ( -368 48 24 ) shared_ex/base1a 0 0 0 0.5 0.5 134217728 0 0
+( -240 48 104 ) ( -368 48 24 ) ( -240 48 24 ) shared_ex/wall_panel7 64 36.5714111328 0 -0.25 0.2187499851 134217728 0 0
+( -240 40 104 ) ( -240 48 24 ) ( -240 40 24 ) shared_ex/base1a 0 0 0 0.5 0.5 134217728 0 0
+( -368 40 104 ) ( -240 40 24 ) ( -368 40 24 ) shared_ex/base1a 160 0 0 -0.5 0.5 134217728 0 0
NetRadiant just rewrote some lines to replace 0.0
with 0
. I have seen similar behavior in the past with 0
rewritten as -0
or the other direction.
We have to check there is only one representation of the same number, at least for zeroes. We have the same problem with angles when a negative value is similar to a positive one but it looks like NetRadiant does not swap them randomly. The fact is NetRadiant swap randomly the zero
representation and adds a lot of noise to commits.