Mantis - Squeak
Viewing Issue Advanced Details
2168 Graphics minor always 11-10-05 10:32 01-31-06 16:32
Herbert Koenig  
closed 3.9  
none 3.9  
0002168: Point>>theta uses single precision constants in the special case and sends in the general case tan := y / x . dir := tan ar
Point>>theta uses single precision constants in the special cases of zero x-coordinate while it uses sends instead of constants in the more general case of negative x.

I've attached a fix using full precision constants and constants instead of sends. There also is test for precision.

I've checked the speed improvement manually and it is considerable.
If someone can tell me how to write a proper test for the speed improvement, please do so!
Verified in 3.7, 3.8 and 3.9

Used the fix in 3.7 but ran the tests only in 3.8

From a mathematical point of view I think this bug is severe but I'm just an engineer and for my field of application I consider it just ugly but good enough.

So feel free to change the severity. [^] (1,065 bytes) 11-10-05 10:32

12-07-05 01:09   
Thanks for the fix; it's been integrated. BTW, for the next bug report, please use multiple attachments instead of a zip file. It's a *lot* easier for people to look and comment if they don't have to go through the extra indirection.
12-09-05 20:30   
Andreas how do we proceed then. You push a mcz into the inbox and send an email to the v3.9 mailing-list
12-09-05 20:39   
All in due time. There is no reason why this particular bug would require an immediate new release of the graphics package. BTW, why did you reopen this bug?
Herbert Koenig   
12-10-05 09:53   
Thanks for including it and I will remember using single unzipped files.
I'm still learning so: Would one single mcz be ok?
As for Andreas comment I see it his way, as long as people only use screen coordinates (even architecture) it will not make any difference.
01-31-06 16:32   
is in 3.9a#latest