(unresolved) 20221011-0997a7c8e6-win_x64-devprev Puzzle: Puzzle 2212 when wiggling, no wiggle takes place, instead some sidechains flap around

Started by Bletchley Park

Bletchley Park Lv 1

Similar issue to the other bug report. Different puzzle though.
Solution uploaded to scientists as "flapping sidechains"

BUG

bkoep Staff Lv 1

Hmm, the backbone of this solution is extremely deformed. Something definitely went wrong somewhere.

Can you say anything about how you made this solution? Did you use any different recipes than usual?

bkoep Staff Lv 1

@"Bletchley Park" Can you share some more information about your system? What operating system and version are you running? Is there anything non-standard about your hardware? Are you using a typical x86 CPU? Or are you somehow running Foldit on an ARM processor?

Somehow, it seems that your client is loading the starting solution with funny coordinates. Specifically, it looks like the (x,y,z) coordinates of all the protein atoms have been rounded to integers (they should be floating point). If you change your View Options to use Stick mode, you might recognize that all of the bonds have funny angles and distances (also, the "Show backbone issues" option should flag virtually every residue). This same thing also seems to cause of the bug you reported with the most recent devprev puzzle.

This is really strange, and unfortunately we haven't been able to reproduce this on any of our development machines. What's more, we are not seeing this kind of thing in solutions from any other users. Do you have multiple Foldit installations, and do you see this behavior on all of them? If you download a fresh install of Foldit, does the behavior persist?

Bletchley Park Lv 1

Bkoep,

  • this is the same machine with the same software I have been using for several years to play thousands of foldit puzzles.
  • Only new thing is the Visual Studio library I needed to install for foldit 64-bit
  • I have a clean new installation of Foldit-64 in a different directory
  • I ran only one instance of foldit when reporting these bugs
  • Apart from that, I have run multiple clients on the same machine, each in their own track, with older client versions for many years without problems.

As I mentioned in my other comment, after resetting the puzzle the scores seem ok and I can use the puzzle. So seemingly in the initial loading something gets corrupted. I have deleted the files in the default directory and reloaded the puzzle, this time it loads fine.

Maybe there is a new routine that in the initial download, unpacking and setup uses a variable that is declared as integer instead of float or double ? Maybe a library function that is evaluated in a different order or evaluated without braces ( ) ? A backward division (a\b) instead of a forward (a/b) may also be a tricky bug.

This compatibility issue with MSVC might be interesting as well:
https://devblogs.microsoft.com/cppblog/microsoft-visual-studio-2022-and-floating-point-to-integer-conversions/

jflat06 Staff Lv 1

Hey @"Bletchley Park", do you know which library you installed to get 64-bit working? I am wondering if the version you installed is a slightly different 'flavor' that may be causing issues.

LociOiling Lv 1

Not sure if it's related, but here's a crash with stuff like this:
core.kinematics.AtomTree: {0} [ ERROR ] Atom3 position -nan(ind)-nan(ind)-nan(ind)

That's after running a ligand tweaking recipe overnight.

No sign of problems on restart.

log.txt