Lots of spurious tiles get loaded when changing base layer to one with a new projection

[jonblower]

I have an OL2.6 application that uses base layers in different map projections. The map is initially constructed with a base layer in lat-lon (EPSG:4326). If I change the base layer to a layer in polar stereographic projection (EPSG:32661) the map loads a whole load (up to 40) of spurious tiles. This effect can be seen by loading the attached example web page and using Firebug (or similar) to examine the outgoing GetMap requests. The spurious tiles have a bounding box that is inappropriate for the current zoom level and map extent (this shows up clearly because the extents for 4326 and 32661 are very different in their native units).

The problem appears using Firefox 2 on Windows XP. It is likely that the problem also occurs on other systems but I have not tested this.

Registering an event listener for changebaselayer does not help: it seems that the spurious tiles are loaded anyway.

[jonblower]

HTML file that shows this bug in action

[edgemaster]

I'm interested in this due to my work on #1249

[elemoine]

I attached a quick-and-dirty patch which fixes the problem seen in the testProj.html example. The patch needs rework, it's just there to show where in the code the bug lies.

[tcoulter]

@jonblower,

I applied your patch, and I'm having trouble seeing the fix. From my end, using testProj.html, switching the base layer does cause spurious tiles to be loaded; however, with the patch, it still seems to load spurious tiles.

I'm not wholly knowledgeable on projections and how they affect base layers -- someone else may need to take a look.

[jonblower]

Thanks tcoulter, but it wasn't my patch, it was Eric Lemoine's - he did say that it was just to show where the bug was rather than a proper fix.

[edgemaster]

Since I want to try and get #1249 into 2.7, I'll look at this while I am working on it this coming week.

[euzuro]

Moving to 2.8 for lack of activity

[crschmidt]

Not touched in 2.7, bumping.

[elemoine]

with patch-1574-r9779-A1.diff setBaseLayer reprojects the center of the map if the projection of the new base layer isn't the same as that of the old base layer.

I actually don't feel confortable with having this in OpenLayers, especially because the reprojection may not possible, for example [-180, -90, 180, 90] (EPSG:4326) cannot be transformed in EPSG:900913 coordinates, because 900913 does not cover the poles. Reprojecting coordinates isn't something OpenLayers currently does internally (for its own needs), and I'm not sure this is something we'd like to change.

Anyway, I thought I could attach my patch to this ticket, for reference.