puzzle picture
BETA Symmetry Debugging
Status: Closed

Summary

Name: BETA Symmetry Debugging
Status: Closed
Created: 10/02/2014
Points: 0
Expired: 12/09/2014 - 23:00
Difficulty: Intermediate
Description: We're investigating a scoring bug that will cause a crash in symmetry puzzles. Unfortunately, this bug affects gameplay very infrequently, so we've had difficulty reproducing it during testing. Play this puzzle to help us reveal the bug! Specifically, we'd like you to share solutions that reliable crash the game (i.e. "Loading this solution and doing [some action] causes a crash every time").
Categories:

Top Groups

RankGroupScorePoints
1Void Crushers14,5470
2L'Alliance Francophone14,2650
3Contenders14,2010
4Anthropic Dreams14,1760
5Go Science14,0460

Top Soloists



Need this puzzle? Log in to download.  

Comments

jeff101's picture
User offline. Last seen 3 days 9 hours ago. Offline
Joined: 04/20/2012
Groups: Go Science
bkoep told me the following

bkoep told me the following about this puzzle:

"In this devprev puzzle, if Foldit detects the bug, it will send us an error report before crashing."

Joined: 07/21/2013
Groups: Beta Folders
Why not all the time everywhere?

Is this true for DevPrev or DevPrev and Main Foldit?

Why not for all puzzles?

Does Foldit detecting a crash slow things down?

jflat06's picture
User offline. Last seen 22 hours 2 min ago. Offline
Joined: 09/29/2010
Groups: Window Group
.

Crash reporting has been a feature for all clients for a while now, and is enabled on all puzzles. It does not slow down the client.

When a Foldit client crashes, it compiles a crash report that is saved out to disk. The next time the client is started, it checks for these crash reports and sends them to the server.

The crash reports give us some basic information such as who crashed, what puzzle they were running, what script (if any), etc. Developers can see these crash reports on the puzzle page.

spdenne's picture
User offline. Last seen 41 weeks 3 days ago. Offline
Joined: 10/01/2011
Groups: Void Crushers
Did you get one from me just

Did you get one from me just now from a crash 2 hours ago?

jeff101's picture
User offline. Last seen 3 days 9 hours ago. Offline
Joined: 04/20/2012
Groups: Go Science
Crash Reports:

Are debug.txt files copies of the crash reports? If not, are the crash reports saved on our machines somewhere else? Where can they be found?

bkoep's picture
User offline. Last seen 10 hours 2 min ago. Offline
Joined: 11/15/2012
Groups: None
Yes

When the client crashes, a report is written to debug.txt. When a new client is opened, it checks for the debug.txt file and uploads the report to the Foldit server.

jeff101's picture
User offline. Last seen 3 days 9 hours ago. Offline
Joined: 04/20/2012
Groups: Go Science
Crash behaviors:

When one of my Foldit clients crashes on this puzzle,
it creates a debug.txt file and puts a message on the screen like below:

Foldit.exe has stopped working
A problem caused the program to stop working correctly.
Windows will close the program and notify you if a
solution is available.

The crashed client also appears all white or at least fogged-over
so I can still read the last score and last command it executed.
The crashed client also disappears when I press the "Close program" button.
Is there a way to control whether the crashed client appears all white
or fogged-over? The fogged-over way may give important clues about a crash.
Perhaps even snapshots can be taken using PrntScr/PrntScrn on the keyboard
and pasting the result into a Windows program like Paint.

bkoep's picture
User offline. Last seen 10 hours 2 min ago. Offline
Joined: 11/15/2012
Groups: None
Not sure

Sorry, I'm not sure exactly what causes the "fogged-over" or "all-white" Foldit window during a crash. I also wouldn't rely too much on information in the Foldit display during a crash. It's most likely the crash was caused by some calculation that took place after the display was last updated.

jeff101's picture
User offline. Last seen 3 days 9 hours ago. Offline
Joined: 04/20/2012
Groups: Go Science
Try the Selection Interface:

On one client that would always look all-white when it crashed, I switched from the Original Interface to the Selection Interface. When it later crashed from the Selection Interface, the crashed client looked fogged-over.

jeff101's picture
User offline. Last seen 3 days 9 hours ago. Offline
Joined: 04/20/2012
Groups: Go Science
Filters on this puzzle:

This puzzle has several filters:

(1) There is a "Residue IE Score" that so far always seems to "pass", giving "no score change".

(2) There is an "Exploration" filter that is "met" when its "Explored" score is under 2.00, but when it is "not met", a horizontal line appears through my overall Foldit score, and this Foldit score does not seem to count for credit.

(3) There is a "Fragment Score" that sometimes has "lost points" like -100, -200, -300, -400, or -600 that seem to reduce my overall Foldit score.

Can you explain what these filters are doing and why sometimes (2) above prevents my Foldit score from counting? Can you cite a wiki or fold.it page that discusses these filters?

Also, is this puzzle a repeat of a previous Foldit puzzle?

Thanks!

bkoep's picture
User offline. Last seen 10 hours 2 min ago. Offline
Joined: 11/15/2012
Groups: None
Filters

These filters are most likely irrelevant to the symmetry bug.

The Residue IE Score and Fragment Score filters are described in most of the recent design puzzles we've posted (e.g. see here).

The Exploration filter simply measures the C-alpha RMSD between the current solution and the starting structure. In this case, it has an upper limit of 2.00 Å, meaning the filter will fail if your solution deviates too much from the starting structure.

This puzzle is a clone of a previous devprev puzzle, and is similar to several design puzzles we posted further back.

jeff101's picture
User offline. Last seen 3 days 9 hours ago. Offline
Joined: 04/20/2012
Groups: Go Science
Exploration filter:

http://foldit.wikia.com/wiki/Conditions shows something similar to the "Exploration" filter discussed above. Both seem to check how far a certain part of the protein has moved from its starting position. One difference is that this wiki page says "Explored 1.25 of at least 1.50", as if it wants the present structure to be less like its starting position, while the BETA Symmetry Debugging puzzle says "Explored 15.26 of at most 2.00", as if it wants the present structure to be more like its starting position. Are these the correct interpretations?

Is there a way to tell which part of the structure the "Exploration" filter is checking? Can you just tell us which part this filter is checking for this puzzle?

Thanks!

bkoep's picture
User offline. Last seen 10 hours 2 min ago. Offline
Joined: 11/15/2012
Groups: None
Exploration filter

The exploration filter compares all C-alphas, including those of the symmetric chains.

jeff101's picture
User offline. Last seen 3 days 9 hours ago. Offline
Joined: 04/20/2012
Groups: Go Science
Checking Exploration's "Show" box shows original structure

Thanks Brian. If I check the "Show" box under the "Exploration" and "Explored" messages, it shows the main monomer's original structure in grey, like a Guide.

spmm's picture
User offline. Last seen 41 weeks 2 days ago. Offline
Joined: 08/05/2010
Groups: Void Crushers
Specific thing you want to test?

Are there any particular scripts or actions you would like to test? Happy to run a script or three but just don't have time to 'play' it.

jeff101's picture
User offline. Last seen 3 days 9 hours ago. Offline
Joined: 04/20/2012
Groups: Go Science
What to test?

On this puzzle, I have been trying to force disulfide bonds to form between the main monomer and its images. I got Foldit to crash a few times in the process.

I recommend going through your usual folding procedure, using each one of your favorite tools and recipes just long enough to see that it works. When you find something that makes Foldit crash reliably, try to isolate what tool or LUA command is behind it. This puzzle allows mutations, so you can try different sequences as well.

I am guessing that the newer tools like idealize, wiggle power, cut points, etc. are more likely to cause crashes. Perhaps certain less-common amino acids make trouble as well.

bkoep's picture
User offline. Last seen 10 hours 2 min ago. Offline
Joined: 11/15/2012
Groups: None
Yes

The biggest difficulty we've had tracking down this bug is the ability to reproduce it consistently. We've only seen it occur in scripts with complex behavior, and even these crashes don't occur every time the script is run.

What would be really helpful is if players could identify a solution that is "one Wiggle away" from crashing the client.

jeff101's picture
User offline. Last seen 3 days 9 hours ago. Offline
Joined: 04/20/2012
Groups: Go Science
Starting Structure is missing certain Amino Acids:

This puzzle's starting structure contains 14 different amino acids but not the following 6:
a=ALA=alanine
f=PHE=phenylalanine
n=ASN=asparagine
q=GLN=glutamine
w=TRP=tryptophan
y=TYR=tyrosine
Perhaps mutating some of this protein's amino acids to one of the above will help a Foldit crash occur.

jeff101's picture
User offline. Last seen 3 days 9 hours ago. Offline
Joined: 04/20/2012
Groups: Go Science
New Foldit Release from 10/21/14 shifts scores down:

The New Foldit Release from 10/21/14 (http://fold.it/portal/node/1998590) shifted some scores for different saved solutions of this puzzle from 13342.806 to 13235.469 (a 107.337 point drop) and 13329.506 to 13220.762 (a 108.744 point drop).

Also, when I tried to load a save from the 10/21/14 New Foldit Release to a client running the previous release, it failed to load the save and instead gave the message:

Unknown residue type 'PRO:NtermProteinFull'.

bkoep's picture
User offline. Last seen 10 hours 2 min ago. Offline
Joined: 11/15/2012
Groups: None
Inconsistent scoring

Yes, solutions are expected to score differently in the outdated client, and the outdated client will not be able to load solutions saved from the latest client.

KarenCH's picture
User offline. Last seen 44 weeks 4 days ago. Offline
Joined: 07/14/2012
Some crashes with a script I just published

I've had some luck crashing the Symmetry puzzle by using BandedWormPairs (just published Nov 5, 2014 at 8:10pm EST) at medium wiggle power. No guarantees the crash I'm getting is related to the Symmetry bug, however, since I can get crashes with it in any high wiggle power game.... (What's unusual here is that I'm able to get them at medium wiggle power).

Karen

KarenCH's picture
User offline. Last seen 44 weeks 4 days ago. Offline
Joined: 07/14/2012
As a note, I've sometimes

As a note, I've sometimes been able to get one of those other puzzles to a "one wiggle away" state, but I've generally considered that a bad thing in the past.... If I can get it there again, I'll do a "share with scientists" of the one that does it.

jeff101's picture
User offline. Last seen 3 days 9 hours ago. Offline
Joined: 04/20/2012
Groups: Go Science
Using math.randomseed() helps give more repeatable results

Using math.randomseed(x) with the same value of x each time should give the same sequence of random numbers. This should help Recipes give more repeatable results, thus helping identify what conditions will reliably crash Foldit.

http://fold.it/portal/node/1998681 discusses this more.

spmm's picture
User offline. Last seen 41 weeks 2 days ago. Offline
Joined: 08/05/2010
Groups: Void Crushers
maybe repost or close?

I now have some time to play with this but the scores are well out of whack - I am playing devprev latest version. The scores are way out of alignment.

my Restore Very Best and Restore Credit best is 13807 - and in game scoreboard shows my credit best as 13913.
SO a bit tricky to actually play it.
I note also that Banded Worm whilst a nice script does crash - usually in the middle of the night or when I have left for work. I have used it on medium and high wiggle power.
I don't use it much any more.

Not very helpful sorry :(

spmm's picture
User offline. Last seen 41 weeks 2 days ago. Offline
Joined: 08/05/2010
Groups: Void Crushers
debug.txt on Mac?

foldit has been crashing quite a bit - not a skerrick of debug.txt on my mac

Get Started: Download
  Windows    OSX    Linux  
Windows
(7/8/10)
OSX
(10.7 or later)
Linux
(64-bit)

Are you new to Foldit? Click here.

Are you a student? Click here.

Are you an educator? Click here.
Search
Only search fold.it
Recommend Foldit
User login
Soloists
Evolvers
Groups
Topics
Top New Users
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