Sketch: Fix algorithm seeking for missing coincidences in Validate

[0penBrain]

https://forum.freecadweb.org/viewtopic.php?f=8&t=62936

[AjinkyaDahale]

Isn't PosIdSketcher::PointPos now an enum class?

[0penBrain]

@AjinkyaDahale yes it is

[AjinkyaDahale]

Then is < appropriate here? Also, if I understand it correctly end=2 and mid=3, which could be troublesome.

[0penBrain]

Not a problem. 'LESS' comparing functions should just ensure a transitive ordering in all cases, and that when 2 equal objects are compared, it returns false in both argument orders. ;)

[freecadci]

for feature branch PR_6393. Pipeline 474829595 was triggered at d20c44b. All CI branches and pipelines.

[wwmayer]

The PR causes a crash when using chrisb's sketch example. For more details see above comments.

[0penBrain]

The PR causes a crash when using chrisb's sketch example. For more details see above comments.

Actually the crash comes from the VertexID_Less comparator (which probably is called with an empty thing). Debugger is warming up. Thx for review.

[0penBrain]

@wwmayer Should everything be OK now. I tested on more test cases (actually I missed the simplest ones, which actually were the most significant).
Let you edit the title of the PR when you think it's right. ;)

[freecadci]

for feature branch PR_6393. Pipeline 480439209 was triggered at 1b65c81. All CI branches and pipelines.

[wwmayer]

Let you edit the title of the PR when you think it's right. ;)

I will have a new look in the next couple of days.

[wwmayer]

Works as expected now. Here is the mandatory unit test to see that current master will fail: 6e4c618

[0penBrain]

Here is the mandatory unit test to see that current master will fail: 6e4c618

Please next time just let me know that a test is expected. Actually I didn't add it because I (wrongly) was pretty sure that this function can't be called from Python. I'll add another one that allowed me to find another flaw in my algorithm. ;)

[wwmayer]

Please next time just let me know that a test is expected.

A unit test is basically always expected. Although Qt has a test module to test GUI intersections I don't expect a function for this kind of stuff.

Actually I didn't add it because I (wrongly) was pretty sure that this function can't be called from Python.

I didn't know about the exposed function either and so I simply searched in the code base. If there weren't one I would have written a wrapper for it.