MacOS 64 odd behavior

Case number:954892-2005449
Topic:Developer Preview
Opened by:actiasluna
Status:Open
Type:Bug
Opened on:Thursday, June 28, 2018 - 23:59
Last modified:Monday, August 27, 2018 - 23:13

Proteins jump about on the screen when working, and eventually disappear from sight.
Reproducible... wiggling and manipulating them particularly using scripts makes the protein "disappear"
Hitting Q to center only works for a while... will recenter but the protein moves farther and farther into the "distance" until you can no longer see it using Q or zooming in.
Entertaining, but not good for the play. :-)

(Thu, 06/28/2018 - 23:59  |  25 comments)


Hanto's picture
User offline. Last seen 5 days 20 hours ago. Offline
Joined: 05/10/2008
Groups: None

Thats been happening since I started Win foldit. Center protein and use your mouse wheel to ZOOM in heartily I might add.

jeff101's picture
User offline. Last seen 15 hours 59 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

I think the centering problem is worse for symmetry puzzles.
It would help if the Home and Q buttons would center the main
monomer on the screen. I think right now they center on the
screen some other point that lies between the monomers.

https://fold.it/portal/node/2004919#comment-36442
https://fold.it/portal/node/2004908#comment-36407

brow42's picture
User offline. Last seen 19 hours 18 min ago. Offline
Joined: 09/19/2011
Groups: None

A single band to space should prevent this and will not affect the folding.

brow42's picture
User offline. Last seen 19 hours 18 min ago. Offline
Joined: 09/19/2011
Groups: None

If mean, if Q is sending the protein into the distance without wiggle, or you have it pinned in space, then that's a problem. Please check.

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

With wiggle and a space band it still leaves the screen.

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

It appears to be moving "the whole space" ... the in-space end of the band moves, and the protein is pulling away from it at the same time.

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

And clarifying... Q is the only way to recenter the protein but that doesn't work consistently. Q won't center it after the protein (and "space") become no longer visible on the screen. A note bubble (which might be a clue) may indicate which direction the protein "went" because I noticed one on screen first try... but not sure, as I could never figure out how to recover the view without reloading the puzzle or reloading a shared protein.

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

Still bouncing around in the space.... loaded latest update and with a space band here's an example....(if I can load the image).
If there are clashes it seems worse.
Also, it is quite slow and without a filter toggle is annoyingly slow (about 2x slower than the current 32 bit version)
(edit) - definitely better than previous as once protein stabilizes in handholding it doesn't jump around as much.
(still first hand-folding stages are dodgy)

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

Without the action "clock" and point monitoring in score when you are working, you also can't tell what effect shaking or wiggling is actually having on the structure. Response is too slow. Don't like this at all for handholding particularly early on. With the slowness it's quite frustrating.

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

The protein is more stable, doesn't seem as prone to completely disappearing. But the responsiveness without the "filters off" toggle function impedes fine control of what you're doing to the protein (seems you see every fifth frame when you try to move, is jerky). Combined with the above two comments, I think this release still has a long way to go to beat the 32bit version. Definitely not a fan of the no timer on the action "clock", (which came back when I went in and restored my advanced GUI in the View, by the way, so I will start another with advanced GUI view (which should have loaded with my options.txt file but did not, apparently)) the removal of the ability to toggle filters , and no way to see how your moves are affecting the overall score as you work (possibly due to the basic GUI use early on.).

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

Also, 64bit client doesn't appear to want to minimize. Don't know if this is by design but typically I can minimize a client that's running a script and this isn't letting me.

joshmiller's picture
User offline. Last seen 19 hours 13 min ago. Offline
Joined: 09/08/2017
Groups: Go Science

Hi all,

Thanks for the detailed report. Wanted to give an update on the dev end of this:

We are able to reproduce the bug and are investigating. Thank you as always for your patience.

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

New Odd behavior for the latest Devprev 64 bit Mac version.
At CI 1.0 all seems well. However, if you wiggle the segment at anything less than CI 1.0, the protein "decomposes"... the backbone disintegrates into a mass of backbone issue bubbles and the protein falls apart.

I brought in the Aflatoxin puzzle to see how it acted and it completely fell apart with the locked segments floating out into space and the aflatoxin losing its moorings and also flying off. It stays within the puzzle space but doesn't eventually resolve.

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

It doesn't display this behavior with the current sketchbook puzzle but it does still move (zoom) the protein in space and off screen. Q doesn't recover it... which is what I think was causing the Aflatoxin to fly apart.
Is the Mac version rolled back to the old one that made proteins effectively disappear?

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

So, on the sketchbook puzzle with one band "tether" here's what happened when it was wiggled at .08 CI... instead of compressing as one might expect, the protein starts to pull apart... moving around on the screen and pulling against the band.

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

I shared both examples with scientists. Good luck finding the invisible one!

LociOiling's picture
User offline. Last seen 8 hours 40 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders
NinjaGreg's picture
User offline. Last seen 11 hours 46 min ago. Offline
Joined: 05/21/2010
Groups: Go Science

Just downloaded today's 64-bit release, and tried it on Puzzle 1558, starting from scratch. Did a shake, select all, "5", shake, wiggle. Protein hops around, but q key brings it back. Then added a single band, reduced strength to 0.25, and wiggled, and the protein disappeared, never to come back. Running a script on it (QuakeR, by Gary Forbis) crashed the client. Will try adding a band to space to the protein, see if that helps.

NinjaGreg's picture
User offline. Last seen 11 hours 46 min ago. Offline
Joined: 05/21/2010
Groups: Go Science

Adding the band did help, sort of. I reset the puzzle, shake, select all, "5" to create helices, shake, add band to one end, wiggle. Bounces all over (even the end of the band in space), and deforms the helices, but I could still see it. Froze the helices, wiggled, it bounced around but not so much. Banded to pull them parallel, again bouncing, but stayed visible.

Completely fresh install, btw.

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

The main 64 bit client release still has major issues, still "deconstructing" the protein into a snarled ball of nothing. The protein also still tries to leave the space.

Changed to devprev which is a bit more stable but still wonky, particularly when first working with the protein. Symptoms include bizarre stretching of segments during wiggle, significantly at LWP, less after working the structure out for a while. Oddly the 64 bit also seems slower than the 32 bit main. Toggling off "pulse while working" at least makes it *look* like it's working more normally. Seems like rendering and spatial controls are still a little iffy (but I have no idea what governs this in foldit so I'm just making a semi-educated guess here.)

Foldit (like other GPU intensive games) is really a CPU and GPU hog even on my iMac Pro...which I "invested" in for better processing and a much better cooling system, after watching my 2015 Retina 5k iMac. With the 64bit release in particular, the system freezes on the old iMac were closer and closer together (Macs are designed to stop processes if machine overheats past tolerances... something it took me about a year to figure out.)

actiasluna's picture
User offline. Last seen 5 hours 43 min ago. Offline
Joined: 03/05/2015
Groups: Gargleblasters

Interesting. I just updated both Silverlight (to solve a streaming video problem) and Flash alerted me that the PPAPI "Pepper" plug in version needed update. After the Flash update I restarted my clients, and now the GPU isn't getting worked as hard.
But, when I opened a 64 bit client(running High Sierra 10.13.6) no difference, still leaps about. Seems inclined to (pun) drift abruptly up to the left corner of the screen if that means anything. Disappeared and eventually (after about 100 scrolls of the mouse) recovered the protein, it had gone "back" past the point it was visible after only about 30 seconds of jumping off screen.

Other issue with the 64 bit (main OR devprev) is that the game won't minimize, which helps as it cuts GPU load by about 80 percent.

phi16's picture
User offline. Last seen 1 day 18 hours ago. Offline
Joined: 12/23/2008

Using version 20180725-d14dae34c3-macros_x64 with an iMac system 10.13.6

I'm seeing the same problems as described above. Protein jumps off screen and Q cannot be used to recover it. I haven't done much experimentation with it since I can't get passed this point when the protein jumps off screen. If its related to having bands, I had bands.

phi16

phi16's picture
User offline. Last seen 1 day 18 hours ago. Offline
Joined: 12/23/2008

There is another bug I see with the new build. It reads "game dynamic" instead of "File". This seemed to have been solved a while back but I see it again with this version. Doesn't seem to affect anything but I never did like this non-mac look/feel.

phi16

NinjaGreg's picture
User offline. Last seen 11 hours 46 min ago. Offline
Joined: 05/21/2010
Groups: Go Science

Hi, I did an "upload for myself" of puzzle 1565 that demonstrates a disappeared, non-Q available solution. I'll attach the log.txt file too, in case that helps.

NinjaGreg's picture
User offline. Last seen 11 hours 46 min ago. Offline
Joined: 05/21/2010
Groups: Go Science

Using build 20180725-d14dae34c3-macos_x64, saved protein is named "DISAPPEARING_PROTEIN_1565".

Sitemap

Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, RosettaCommons