Automated issue posting is cluttering the GitHub Issues section

[XX-Yin]

A lot of issues were automatically posted on the GitHub Issues based on the automatic logging, which makes navigation very difficult. I also posted a similar issue before https://github.com/AllenNeuralDynamics/dynamic-foraging-task/issues/949

Issues associated with this auto posting are:

• Most issues are redundant
• The logging system produced nonsense errors. For example, closing GUI in a wrong order (e.g. bonsai first then GUI first) can generate an issue.
• The logging system has overstated some issues as errors, which would ideally be categorized as warnings (one example is the screenshot shown before; another example).

@hanhou @alexpiet what do you think?

[alexpiet]

• If things are overstated as errors instead of warnings, we should change how we log them in the code.
• For nonsense errors, like closing the GUI in the wrong order, we are marking them as "wont fix" which will prevent the error from re-opening.

[XX-Yin]

The primary source of clutter is from the redundant isses. You can check and most of these issue are of this type which makes it very difficult to browse the issue section.

[ellahiltonvano]

@XX-Yin a helpful thing to do might be to use the project board, rather than the issues tab. I have issues organized in the milestone view so they are easier to view

At this point, its important that we are logging these issues so Alex and I can get a sense of where things are going wrong in the code.

[XX-Yin]

@XX-Yin a helpful thing to do might be to use the project board, rather than the issues tab. I have issues organized in the milestone view so they are easier to view

At this point, its important that we are logging these issues so Alex and I can get a sense of where things are going wrong in the code.

@ellahiltonvano The project board is helpful.

[XX-Yin]

I think we can avoid duplication of issues by:

• Treating the same issues occurring on different rigs as a single issue.
• Treating issues traced to the same line of code as a single issue.
• Avoiding duplicate posts if an issue, based on the above criteria, already exists in the GitHub issue section.

The first two criteria seem relatively straightforward to implement, but I’m unsure how to fetch GitHub issues using Python code, which is a prerequisite for meeting the third criterion.

[alexpiet]

• Treating the same issues occurring on different rigs as a single issue.

This already happens

• Treating issues traced to the same line of code as a single issue.

Code lines are ignored because as code updates happen code lines change. Instead issues are grouped by the issue text.

• Avoiding duplicate posts if an issue, based on the above criteria, already exists in the GitHub issue section.

This already happens. If the issue already exists, my code comments on the original issue.

[XX-Yin]

• Treating the same issues occurring on different rigs as a single issue.

This already happens

• Treating issues traced to the same line of code as a single issue.

Code lines are ignored because as code updates happen code lines change. Instead issues are grouped by the issue text.

• Avoiding duplicate posts if an issue, based on the above criteria, already exists in the GitHub issue section.

This already happens. If the issue already exists, my code comments on the original issue.

For the second point, it seems that most duplicates are from the same GUI code with the same code lines (example shown below). So, adding this rule will remove most of the duplicates.

Additionally, we can check line number and error information, or code in that line and error information.