Bugs in Transform?

[ljbade]

After poking around trying to figure out #1468 I have come to the conclusion that parts of the code in Transform that iOS has stopped excercising for a while has become stale.

There are some really strange bugs I have found particularly with Map::setZoom and Map::setLatLng which only Android has been using recently.

Looking at iOS it uses Map::setLatLngZoom in combination with Map::getZoom and Map::getLatLng to implement what used to be done with Map::setZoom and Map::setLatLng. Also none of the C++ test cases seem to use either of the those functions.

I can change Android to switch to the same code that iOS uses, but that will only cover up the bugs in the C++ layer. I might do this as a stop gap to get a working Android app until the bugs in the C++ side have been fixed.

Solutions I can think of:

• remove Map::setZoom and Map::setLatLng since no one using it anymore
• fix the functions and create test cases that cover all features of Transform to stop future bugs
• leave Map::setZoom and Map::setLatLng alone and just leave Android and iOS using Map::setLatLngZoom

/cc @incanus @bleege @jfirebaugh

[bleege]

/cc @kkaefer

[jfirebaugh]

If there's a bug in setZoom or setLatLng we should certainly fix it (and add tests). I'm hesitant to jump straight to that conclusion though. Have you been able to trace through and see where things go wrong?

[incanus]

Does order of calling matter? e.g. set center, but then set zoom and it loses/invalidates the center?

[ljbade]

@incanus I was thinking along those lines, but I changed to using only setLanLngZoom and it workse slightly better but it is still rather broken.

[jfirebaugh]

We determined this was due to problematic floating-point compiler options.

[bleege]

We determined this was due to problematic floating-point compiler options.

Awesome! So this means that compiler options have now been adjusted and things work on ARM again?