[squeak-dev] #07212 flaps render *properly* during StandardSystemFontsTest
Hi Timothy,
:-) I am really happy you posted this. It means someone else notices the small symptoms of big bugs, too.
If you run some of the font tests. The fonts will change and then be restored the defaults. Changing the flap fonts causes the flaps to be laid out again. This may either accidentally and temporarily cure the off-by-one/roundoff error that affects the flaps. There are some other bugs in flaps as well. I made repairs for them but I don't know and rather doubt whether they actually got into 3.10.2.
If you view sub issues of #7212 on mantis you can see what else has been discoved. (http://bugs.squeak.org/view.php?id=6486 is one of the main reports.)
Yours in curiosity and service, --Jeorme Peace
***
Mr. Timothy C. Johnson tjohnson at iwu.edu Sat Nov 29 01:32:39 UTC 2008
Hi,
It is known that flaps render improperly in 3.10.x -- there is a ~1 pixel dead zone on some of the flaps. But I just ran System-Support-Tests>>StandardSystemFontsTest and something it did made the flaps render properly, all of a sudden. Any idea why?
- timj
***
On Dec 4, 2008, at 12:07 AM, Jerome Peace wrote:
:-) I am really happy you posted this. It means someone else notices the small symptoms of big bugs, too.
Yes, I think small things like this make a difference for new people discovering Squeak. Glitchy graphics = perceived glitchy system = new users walk away. Also, like you said, they are often the bubbles at the surface of the pool which indicate the monster lurking in the deep :)
If you run some of the font tests. The fonts will change and then be restored the defaults. Changing the flap fonts causes the flaps to be laid out again. This may either accidentally and temporarily cure the off-by-one/roundoff error that affects the flaps. There are some other bugs in flaps as well. I made repairs for them but I don't know and rather doubt whether they actually got into 3.10.2.
You are right. The flaps being laid out again is what appears to fix the rounding error. I have read some of the other Mantis reports on this but I don't have any insight into them.
Have your fixes been acknowledged/reviewed by others for inclusion in release?
FWIW, I noticed that when I took the corrected image from my Mac and used it on a PC, the flaps were rendered improperly again. Re-running the test fixed the flaps. I then took the same image *back* to the Mac, and the flaps were rendered improperly once more. Again, the test fixed the flaps. What could be happening between platform shifts that could be resetting these values? More rounding errors? (I wonder, if Morphic was originally implemented on a non-rounding renderer/bitblt system, why rounding was added much later...)
Thanks, TimJ
squeak-dev@lists.squeakfoundation.org