-
Notifications
You must be signed in to change notification settings - Fork 38
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
GUI stops updating #763
Comments
Let me know if anyone manages to reproduce this reliably ... it happened to me quite often trying to build a model, but I'm not managing to reproduce it on purpose. |
Further console output examples
|
It seems it is required to move objects around, but still can't reproduce it reliably. |
@PLindes Are you using the |
I posted that issue. He's using the |
Is this related to Issue #759 ? |
Don't think so |
This is not related to the server refactoring. It also happens on
|
I am finally able to consistently reproduce this with one model file I have (not sure if it work for any). In the model file the following steps are required:
I can go back in versions and still reproduce this in version 0.1.3, probably even older versions (wanted to document this before testing further). |
ok, I think, I've got a rough idea what's going on. To understand that, we need to look at how the position is stored on the Python side: Each Nengo object is assigned a uid (think of it as a name), this uid is mapped to the actual Nengo object, that actual Nengo object (or it's memory address) is used to map to the position parameter in the config. When the ensemble is initially created, everything gets assigned and mapped properly. But when the code is edited than somehow the Nengo object changes (it's memory address will be different), probably due to the recompilation. The uid mapping is updated accordingly, but not the config with the position. So the new Nengo object due to recompilation will have a position config of |
A potential quick fix might be to set the |
Personally, I am not familiar enough with the code base yet to propose what a proper fix would be right now. I have to figure out in detail, how stuff gets updated and objects get related to each other on compilation. |
With these details, I can take a shot at the bug. Alternatively, if you want to do some reading, the Flow Docs might help. Do you know where to find those? |
I don't think there is enough details in the Flow Docs. Feel free to take a shot at fixing the bug. I'm probably also looking further into it until I go to dinner. |
Though #767 might make it a bit tedious. |
Actually the problem might be not so much that the config is not updated for the new object id, but that it is completely reloaded anyways? Where there any changes to when the config gets written to disk? |
We don't have anything in Nengo right now to copy the config of an object, do we? (@tbekolay?) |
Changes to what? |
to when the config gets written to disk |
I have a branch where it is (I believe) possible to pickle configs, as I needed it to parallelize some model. It might be somewhat out of date though. https://github.com/nengo/nengo/compare/pickle-config |
This is actually probably related to another error some of the users were getting where if they ignored this error long enough there ends up being too many files open which causes another error. |
Not sure. How would this be related? On June 8, 2016 6:17:18 PM EDT, Sean Aubin notifications@github.com wrote:
Sent from my Android device with K-9 Mail. Please excuse my brevity. |
It's just a hunch. I think the problem we're experiencing has to do with On Wed, Jun 8, 2016 at 6:52 PM, Jan Gosmann notifications@github.com
|
@jgosmann how did you isolate the bug to the fact that the config is being changed in the wrong way while the code is being written? |
New reduced steps:
Example resulting config:
|
New reduced steps.
I've narrowed down the bug to |
This issue is a major annoyance for me right now. 👿 |
grrr. I'm looking into it now, and at the very least here's a quick PR that tosses in an exception handler that should mean that everything doesn't fall apart when this exception happens: #842 |
Thank you, @tcstewar! 🙇 🙏 |
with the following exception on the command line:
(note that this is on the
server-refactor
branch, #673, but it doesn't seem to be related to those changes to me)The text was updated successfully, but these errors were encountered: