Hi folks,
Is there any persisting pressing sketcher bug that we should attempt to fix before release?
[Sketcher] Pressing bugs before Release
Forum rules
Be nice to others! Respect the FreeCAD code of conduct!
Be nice to others! Respect the FreeCAD code of conduct!
Re: [Sketcher] Pressing bugs before Release
A lot of issues have been now moved to "out of scope" unless there is an active effort at fixing them. See https://github.com/orgs/FreeCAD/projects/4/views/2.
There are a lot of issues regarding how constraints are not correctly counted. These probably should be handled together by modifying the solver's redundancy check etc. It would be good if we could fix them now but probably we won't have enough time for that. Ideally these should be backported if possible.
One issue that we can look at is https://github.com/FreeCAD/FreeCAD/issues/7520. It seems to be something you would be able to quickly catch.
There are a lot of issues regarding how constraints are not correctly counted. These probably should be handled together by modifying the solver's redundancy check etc. It would be good if we could fix them now but probably we won't have enough time for that. Ideally these should be backported if possible.
One issue that we can look at is https://github.com/FreeCAD/FreeCAD/issues/7520. It seems to be something you would be able to quickly catch.
My latest (or last) project: B-spline Construction Project.
Re: [Sketcher] Pressing bugs before Release
This regression please https://github.com/FreeCAD/FreeCAD/issues/9208
Re: [Sketcher] Pressing bugs before Release
https://github.com/FreeCAD/FreeCAD/pull/9681jnxd wrote: ↑Sat May 20, 2023 7:11 am One issue that we can look at is https://github.com/FreeCAD/FreeCAD/issues/7520. It seems to be something you would be able to quickly catch.
If you mean popularity contest algorithm issues (conflict/redundant undetected or wrongly classified), or constraint getting degenerate in a corner case (DoFs do not correspond to actual DoFs)?
For the first, we need to increase the functional test coverage first. I have opened this as an issue:
https://github.com/FreeCAD/FreeCAD/issues/9683
For the second, it is best to tackle these early in the development cycle. Often the solution is to provide a non-zero divergence for the parameter that is zeroed due to combination of constraints (often 1e-13, which would need to be made official by not hard-coding it, but using a named constant). This tiny divergence has not interfered with convergence during DogLeg, yet it allowed the system not to degenerate during QR decomposition. If this is not a general solution, we would need to write two functions per constraint, one for DogLeg and another one for QR decomposition, which I am wary to do as it is much more code to maintain only because of some corner cases.
Re: [Sketcher] Pressing bugs before Release
Thanks for fixing!abdullah wrote: ↑Sun May 28, 2023 4:56 amhttps://github.com/FreeCAD/FreeCAD/pull/9681
If you mean popularity contest algorithm issues (conflict/redundant undetected or wrongly classified), or constraint getting degenerate in a corner case (DoFs do not correspond to actual DoFs)?
For your question, I think most of the issues reported fall under the first category. By early in the cycle (2nd case), do you mean we move it for a future version?
My latest (or last) project: B-spline Construction Project.
Re: [Sketcher] Pressing bugs before Release
Welcome.
I am not against fixing (or trying to fix) the ones in the first category before release. However, we do need to increase the functional test coverage before we touch that function. I merged lately some code to generate the Python statements to create a sketch from the sketch itself, this is very useful to create these unit tests:
https://github.com/FreeCAD/FreeCAD/issu ... 1566063410
@Jolbas agreed to look into making some of these unit tests, but he cannot do it in the short term. If you have time now and want to help there, I am sure he will welcome some help.
In fact, it would be great to look into closed issues for popularity contest and create as many tests as possible.
Regarding the degenerate cases, the problem is that we may fix them and generate another issue in another corner case. The likelihood of finding that other corner case before release is very low. The impact of the other corner case may be bigger than what we are fixing (bad is that 1 DoF is not detected, worse that 1 extra DoF is reported).
So both issues have a similar problem, with the difference that for the first we have quite some sketches to do good tests to manage the risk. For the second, it is difficult to identify the "other corner case" that may be affected, and thus writing meaningful tests. This is what motivates me to have a different approach to both.
Re: [Sketcher] Pressing bugs before Release
Unfortunately I don't think I'd have the time to do this before the release, almost by definition. The priority is to clear out any blocking issues, mainly the toponaming ones. So any time spent on an issue not in the list would only delay the release.
My latest (or last) project: B-spline Construction Project.
Re: [Sketcher] Pressing bugs before Release
My bad. Sorry!
Sure. We can do it later on.