Very bizarre (and specific) bug in Golly 2.6

[Sphenocorona]

Steps to reproduce:

1. Set the algorithm to QuickLife. This bug only happens on QuickLife.
2. Make sure the zoom scale you are using is between 1:32 and 1:4 (inclusive - 1:32 is the most likely to show it).
3. Put the row with Y coordinate -896 as the very bottom visible row on-screen.
4. Draw any still life so that it has cells in the bottom-most row of the grid (a.k.a. -896)
5. Press space to move forward a generation.

Why does QuickLife act so weird around the number 896?!?

Note: This also happens with Y=+896, as well as X=896 or X=-896 on the left side of the screen (These row/column values may show the effect for fewer zoom settings). The right side and top side show no symptoms.

[flipper77]

Yeah, I get the same problem myself with Windows 7 64-bit version. Did a bit of testing:

• This visual bug happens on the left and bottom side of the screen(hide scripts/patterns)
• It seems to happen at (+/-)((2^n)-1)*128 coordinates, both x and y axis's for n >= 1
• Place a block so that its top-left cell is at 128 -129 and view at very lower-left corner: get both horizontal and vertical bugs at the same time

I wonder what could cause this, I have no idea how this would occur.

[Andrew]

Yep, I see the same bug. The QuickLife algo uses a stagger-step method that means a block is in a different position (internally) at alternate generations. This is the reason patterns seem to "shiver" at scales above 1:1 (eg. try generating a block at scale 2:1), but there certainly shouldn't be any display problems at scales 1:2 or below.

It looks like we're not calculating the visible boundary correctly at odd generations (but only at certain locations and scales). This bug has probably been around since day 1 of Golly (nearing 10 years ago), so well spotted! I'll probably need Tom's help to fix this.

Thanks for the report and excellent analysis.