fix: 1874 - fixed side-effect crash for final dot / final dash

[monsieurtanuki]

Fixed a side-effect crash for both final dot and final dash.
Still, the display may look funny on final dash mode with PatternFit.extendFinalDash, but that's probably how it's supposed to work (e.g. not well at the end in special cases).

[josxha]

Thanks a lot for providing a pull request this quickly, @monsieurtanuki! I can confirm that this fixes the bug.

[monsieurtanuki]

Thank you @josxha for your review and for having put enough details in the issue (video + log) for a quick fix.

Still a bit worried about possibly slower performances of the new dash/dot/solid algorithm: is there a sort of unit test dedicated to polygon performances?

There may also be minor interferences between culling and dash/dot algorithm: if geopoints are preventively dismissed as "too far away", the dash pattern change (e.g. as starting from a different geopoint).

[JaffaKetchup]

The best way to test performance is to build in release or profile, then use the Polygon/Polylines Stress Test in the example app. Disable "Performant Rendering", and try changing the thickness of the line. Then compare the performance graphs at the bottom (not available on web) to an older version.

[monsieurtanuki]

From data I've just gathered, the previous code is slightly better in extreme conditions, but the current PR is similar in normal conditions.

Optimization tracks (beyond this PR):

• use the "natural" paths for polylines with decent coordinates (below 64M) and switch to the precomputed version beyond
• the polyline cropping I've already coded for osmdroid (cf. [BUG] v7-beta: Polyline becomes distorted & invisible (but remains interactive) when zoomed far in #1864 (comment))

But I would recommend merging the current PR quickly, as it fixes a crash for dashed and dotted

x	latest PR	previous code
optim
no optim

[josxha]

But I would recommend merging the current PR quickly, as it fixes a crash for dashed and dotted

Agree 👍🏼