Incorrect rendering of ways

[jfitie]

Sometimes certain ways seem to render incorrectly. They look like they loop over themselves and fan out. Attached to this ticket are some images to describe the problem, taken on Android using SVN 5197 with pitch set to 10.

screenshot_2012-08-10_2004.png: Highways look normal. screenshot_2012-08-10_2006.png: Pan a little, and ... screenshot_2012-08-10_2006_1.png: Location of screenshot.

I have also seen this happen with railways.

Any ideas as to why this might happen and how to solve this?

[sleske]

Hard to fix this without further information. Any hints on how to reproduce this?

Ideally, please post the map files you used (put them up somewhere for download, as they are probably too large for attaching here in trac), along with information where you got the maps.

[jfitie]

What other information do you need? I have downloaded the map of the Netherlands using the Android port. Location is in 3rd screenshot; but not specific to just this location. I have seen this happen all over the country. Sorry, but I really can't be more specific. I hope the first 2 screenshots clearly show what the problem is (top right of the 2nd one); I have also seen this happen with railways.

[sleske]

Thanks, that's the information I meant :-). I'll download the map and try to reproduce the problem.

[sleske]

Render error at same position

[sleske]

Hm, I cannot reproduce the problem exactly, but I do get similar behaviour. I downloaded a map of the Netherlands from Navit's planet extractor, and moved to the position indicated in this report.

I I switch to 3D view, and zoom in a lot, after some panning I get funny "fanning out" of the colors of the motorway (see screenshot rendererror.png). This is on Linux, using the gtk_drawing_area graphics plugin). I'm not certain this is the same problem, but I'll try looking into it.

[sleske]

​rev. 5417 fixes a clipping error for wide roads at high zoom levels, which caused these roads to be drawn much too wide (possibly with a width of several hundred times the screen width).

Unfortunately, on Linux I cannot reproduce the exact problem you are seeing, and I do not have an Android device to test on. However, the bug fixed

• is in code that runs on all platforms
• causes incorrect drawing coordinates (very large values)
• results in rendering errors on Linux

so it seems likely that it is the cause of the errors you are seeing.

In ​rev. 5418 I also fixed an integer overflow in polygon clipping, which meant the intersection between polygon and screen edge was sometimes calculated incorrectly, leading to incorrect rendering. This is probably not the source of the bug in your screenshot (because there the error is with a street, not a polygon), but it may cause similar rendering errors.

Therefore I am closing this bug. If the problem persists with this fix, please reopen :-).