Activity for High Performance Coder
-
High Performance Coder posted a comment on ticket #1897
Turns out my issue was not related to this ticket after all. It appears that a recent electron upgrade caused a problem with Windows builds canvas blanking out. I've been able to solve the problem by downgrading electron to version 37. -- Dr Russell Standish Phone 0425 253119 (mobile) Principal, High Performance Coders hpcoder@hpcoders.com.au http://www.hpcoders.com.au
-
Seems like version 37 works fine, just not later versions.
-
Upgrade electron
-
I just got this problem: I tried checking them on a Windows builds, and nada! Canvas is completely stuffed, reminiscient of ticket 1897: https://sourceforge.net/p/minsky/tickets/1897/. Funny thing was, these builds were working yesterday, but now anything built after 10th December fails to work, with no error messages on console to indicate what might be going wrong. The last Windows update I had is dated 11/12, but then why would builds I did yesterday, work yesterday, but stop working today. Very...
-
Did you get a chance to try this build out?
-
Implement drag'n'drop support
-
Make RavelCAPI threadsafe
-
See https://www.hpcoders.com.au/ravel-3.18.0.exe . This is a build of that release with the current development environment.
-
That's a lot of commits to go through, since 9th July! Nevertheless, I looked through all commits looking carefully at any that affected code relating to drawing on the canvas over multiple threads. But they were all in code paths that are only active on MacOSX. The only other significant thing was my computer's SSD dying, and having to rebuild the development environment. What I can do is rebuild 3.18.0 with the modern development environment and send it to you to see if it exhibits the hangs.
-
Did you ever get around to checking out older releases to see where the GUI freeze started?
-
GUI freeze
-
Did you ever get around to checking out older release to see where the GUI freeze started?
-
About box popup crashes
-
Problem turned out to be a spurious shell.beep() command in the about box code, left in by the original developers. Removing this line - which basically does nothing useful, and the crashes go away. It is the only instances of this call. There are no reported examples of shell.beep causing a crash in recent electron versions, and its too labour intensive to produce a minimal version demonstrating the bug to file a bug report.
-
For instance - if you can verify that the problem started somewhere between 3.19.0-beta.10 and 3.19.0, then I'll take a closer look at the code changes around mutexes in that range. However, visually, it doesn't look like anything on the Windows code path was touched, only Macintosh code paths.
-
As is typical in this sort of thing, I've been unable to reproduce it, which leaves me completely in the dark. I had the same problem with the MacOSX crashes earlier, but at least I had a stack trace to work on to narrow things down a bit then. If you find an earlier version that doesn't exhibit the problem, and a version that does, that can give clues. Another datum that might be relevant is the Windows version. I'm currently on build 26200.7171 (Version 25H2). It tells me this is the latest version...
-
From the video, it looks like the application is still responding to the mouse, just not redrawing on the screen. Possibly this is a deadlock in the screen redraw process. Do you have any hints as to reliably reproduce this? I've never seen this happen on Linux, of course, so it might be a Windows only issue. Also, any hints as to when this started happening?
-
Add Leap 16 to obsCheck and Ravel build targets
-
Centering Bug
-
Turns out that an upgrade to Cairo internals a few years ago means that rendering to a canvas of size 2e9 pixels squared fails, and has to be reduced to 2e6 pixels each side. I had come across this problem before in unrelated EcoLab work, but the change also needed to be done on one line in the Minsky codebase. The issue lurked in the background, hidden whilst I was using outdated versions of the library on MXE, only to be exposed when my computer's SSD died and the dev environment rebuilt as a consequence....
-
Autodetect presence of webasm support in browser
-
Closing this ticket as stale. Recreate it if there's ever a need for it.
-
Feedback on import form
-
My suggestion of 2024-11-13 was implemented. Closing this as stale.
-
Godley Table Title Bug
-
Not sure how this happened - possibly in a AI agent commit, although nothing seems to have changed in the relevant code for a couple of years. Needed to add a "nameCurrentItem" call before popping up the dialog to collect the new title name.
-
Oh dear - bisect didn't really help. Building 3.18.4 from source code (on both Linux & Windows, Windows requiring adding -lgdi32 to the command line) gives the misbehaving version. Going back in time, current builds of older versions give misbehaving versions right back to when the code is no longer compilable with modern dev tools. What has changed from before to now is that my development system has had to be completely rebuilt after the SSD died. What has changed? Why just about everything, dammit!...
-
Add Leap 16 to obsCheck and Ravel build targets
-
Add Leap 16 to obsCheck and Ravel build targets
-
It is possible to add a flow variable and a parameter of the same name
-
Hang after variable browser close
-
Hang appears to be fixed with a change to mutex handling suggested by Github Copilot.
-
Linked Calipers not working
-
Looks like the second dataset had the wrong row selected for the header row at some point.
-
You've reported this at https://sourceforge.net/p/minsky/tickets/1794/ and https://sourceforge.net/p/minsky/tickets/1787/ and maybe other places. This ticket is about something else entirely. In all the examples you sent of the "3 stooges problem", correcting the miscreant line fixed the problem - I never needed to edit the .mky file directly, nor even save and reload (a somewhat less egregious sin). So I'm at a loss as to what needs to be done.
-
I don't believe it is releated. This has to do with initial conditions being defined in terms of the variable. What you're talking about stems from a misunderstanding of signs between assets and liabilities, I believe. On Thu, Oct 30, 2025 at 06:16:15AM -0000, Steve Keen wrote: This is related to the other bug, when a logic error in defining flows--one that generates A-L-E not equal to 0--pops up in other tables if you delete it in one. A "delete from all tables" command might help in both cases....
-
Maybe solved by moving browser functionality into an HTML window, as per https://sourceforge.net/p/minsky/features/233/
-
Variable browser not functional on MacOSX
-
Variable borwser not functional on MacOSX
-
Cannot recover from mistakenly adding a loop definition
-
Godley Table dropdown not working
-
OK - I gave this task to Github Copilot - it didn't do too bad a job, maybe about a 90% job, which I fixed up, then promptly refined with testing. Only question is whether the location of "Import stock variables" in the context menu is optimal.
-
On the canvas at any rate - obviously there is that support in the Godley popup. Note there may be a solution of adding some support in the context menu. A bit of logic to code in Typescript, but might be a good job to give Github copilot. On Sun, Oct 26, 2025 at 09:10:52AM -0000, Steve Keen wrote: So you have to type stock names in then? No assistance from Ravel in knowing what they are? ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ [tickets:#1890] Godley Table...
-
That's because the triangle wedges are vestigial items left over from when the code supported the Godley popup window. Since Niel's refactor of the Godley popup, they're not needed. It is not feasible to have the canvas drive menus. The wedges should be removed from the edit mode tables. -- Dr Russell Standish Phone 0425 253119 (mobile) Principal, High Performance Coders hpcoder@hpcoders.com.au http://www.hpcoders.com.au
-
Make "Editor Mode" a default in Options
-
See https://sourceforge.net/p/minsky/features/257/
-
There doesn't seem to be any way to remove all vertical/horizontal markers from a plot.
-
Also may solve https://sourceforge.net/p/minsky/tickets/1888/
-
Maybe solved by moving browser functionality into an HTML window, as per https://sourceforge.net/p/minsky/features/233/
-
Stash info from /etc/os-release at build time
-
Finally got this to work, and tested on Linux Mint. So upgrading to ravel should work from version 3.19 for other distros other than our directly supported distros, provided you can get a version of Minsky that runs.
-
Finally got this to work, and tested on Linux Mint. So upgrading to ravel should work for other distros other than our directly supported distros, provided you can get a version of Minsky that runs.
-
Running average (etc.?)
-
Title menu item in Godley popup doesn't set correct item
-
Title menu item in Godley popup doesn't set correct item
-
Actually fixed by Github copilot
-
Correspondence copied from Patreon chat: Thanks - yes I was able to reproduce this issue on both the 3.19 beta and the latest 3.18 release on MacOS. Conversely, this issue doesn't present when running the software on Linux. I'm not able to test this on Windows ATM. Would it be possible for you to raise a bug report at https://sourceforge.net/p/minsky/tickets/ so this gets tracked and fixed. Also, if possible, could you try earlier versions of Ravel to see if this is a regression
-
Hang after variable browser close
-
Stash info from /etc/os-release at build time
-
Calipers not added when first joining group
-
Joined Ravels not working
-
MXE container fails to build postgresql library
-
I have worked around this by building postgresql on a native installation of MXE, then copying the binary artifacts to the docker container.
-
Export to CSV and LaTeX menu items on Godley popup don't work
-
Allow copy/pasting multiple cells into the Godley table popup
-
Again workarounds exist, as this functionality does work when called from the context menu.
-
Export to CSV and LaTeX menu items on Godley popup don't work
-
Low priority, as workarounds exist - eg the context title menu item works correctly.
-
Title menu item in Godley popup doesn't set correct item
-
Ticket moved from /p/minsky/tickets/1830/ Can't be converted: _priority: 5low
-
MXE container fails to build postgresql library
-
Replace all instances of *ngIf, *ngFor with @if and @for
-
I think you'd find it just as much a chore to specify the multiple formats in the file upfront. And Ravel will need to be stricter about applying the formats - oh that format doesn't match, lets try the next one on the list, and so on. The problem is in the data. Currently Ravel has the capability to manage such a mess, but it requires some thought from the user. There's no way of automating it.
-
I think you'd find it just as much a chore to specify the multiple formats in the file. And Ravel will need to be stricter about applying the formats - oh that format doesn't match, lets try the next one on the list, and so on. The problem is in the data. Currently Ravel has the capability to manage such a mess, but it requires some thought from the user. There's no way of automating it.
-
Something went wrong previously with the attachment. This is what it is supposed to be.
-
Not if you set the display format to %Y-%b or %Y-%m. To get that display, you still have the format incorrectly set to %Y-%Q
-
As the software currently stands, you can import the monthly and annual series separately, using formats %m/%Y and %Y respectively, then write out the series as separate CSV file using Ravel's export to CSV feature.
-
Note formats with only %Y %y %m %d %H %M %S and nothing else can handle can handle missing fields. The fields are filled in the order specified - maybe it should be modified to accept y m d h m s preferential order in the event that fewer fields are provided in the input - sort of like %d %m %Y and if that doesn't match then %m %Y and if that doesn't match %Y
-
Date Formats
-
I'm not sure what you mean by "new user". These data files are sufficiently complex, it would take someone who's been around the traps to navigate their complexity. As it was, I noticed you had the display format for Date set to %Y-Q%Q, which is not appropriate for monthly data. So I changed that to %Y-%b. Then I noticed you had somehow reset all the dates to quarters - not sure how you did that, so I had to reimport the LONG_CPI file, using the blank format. This of course adds an extra handle for...
-
Launch from PowerPoint Javascript error
-
I'm going to close this issue then.
-
Connecting to a database causes Minsky to crash on exit on MacOSX
-
Replace all instances of *ngIf, *ngFor with @if and @for
-
Google AI suggests: ng generate @angular/core:control-flow
-
Failed to find shared library for backend sqlite3
-
Thanks Joel. I have now enabled those backends in MacPorts, updated the documentation for setting up a production build, and edited the script that creates the release bundle to include all available SOCI backends. Now, it appears the Windows production build has exactly the same problem, so I'll need to do the equivalent there. Finally, it would be nice to obtain a list of available backends at runtime, and populate the drop menu with just those backends.
-
How is this not a dup of https://sourceforge.net/p/minsky/tickets/1866/ or https://sourceforge.net/p/minsky/tickets/1868/?
-
Failed to find shared library for backend sqlite3
-
Name command on import
-
Only happens if you click on the newly created parameter. So I've created additional logic to handle the case the user does that. If you click on any other canvas item, then all bets are off - the rename just will not happen, as the system has lost sight of what it is that is being loaded.
-
Report actual error when ravel plugin fails to load
-
Upgrade fail
-
Tell that to Microsoft. The cause is a well-known bug^H^H^Hfeature of Windows that it inherited from MSDOS. This only happens if you are trying to install Ravel more than once simultaneously.
-
Save as files should default to same directory as previous used
-
Save as files should default to same directory as previous used
-
Now done - with two different default directories: :models for most things, and :data for imported CSV files.
-
This behaviour has started pipping me enough to do somethign about it. The strategy is to wrap the standard electron dialogs by one that save the returned parent path.
-
Save as files should default to same directory as previous used
-
The error message means that the bindings package within node.js fails to find minskyRESTService.node. I assume that when you launch the Ravel app directly from the Windows desktop shortcut, things work fine, so bindings is finding the node addon. Actually there are no changes between 3.16 and 3.17 in the node area, so the only thing that could have changed is how MSOffice deals with forking a node application. So, if possible, could you verify if the 3.16.19 version still works now. Note I do not...
1 >
