[Tilesathome] Strange rendering issue

[Gregory Williams]

> -----Original Message-----
> From: tilesathome-bounces at openstreetmap.org [mailto:tilesathome-
> bounces at openstreetmap.org] On Behalf Of Gregory Williams
> Sent: 15 February 2009 16:48
> To: Ed Loach; tilesathome at openstreetmap.org
> Subject: Re: [Tilesathome] Strange rendering issue
> 
> > -----Original Message-----
> > From: tilesathome-bounces at openstreetmap.org [mailto:tilesathome-
> > bounces at openstreetmap.org] On Behalf Of Ed Loach
> > Sent: 15 February 2009 10:46
> > To: tilesathome at openstreetmap.org
> > Subject: [Tilesathome] Strange rendering issue
> >
> > There are a number of tiles that have rendered strangely in a
> vertical
> > line between Cambridge and Huntingdon in England:
> >
> > http://www.openstreetmap.org/?lat=52.2911&lon=-
> > 0.0339&zoom=12&layers=0B00FFF
> >
> > I have put them all back into the requests queue, but rendered one
> > manually here to see what happened, and got the attached at z12
which
> > looks a bit wrong to me (I'd have expected the tile to fill the
> image).
> >
> > Any suggestions what may be wrong?
> 
> I wonder whether it could have anything to do with the Greenwich
> Meridian running down the right-hand edge of these tiles?
> 
> As I understand it the tiles are rendered slightly over size (to
> account for long place names, etc.), then clipped appropriately.
> Looking at z12 2047 1349 as an example I see features from each of the
> adjacent tiles, so it's as if the unclipped tile is getting scaled to
> fit inside the normal tile's bounds, instead of being clipped to those
> bounds.

It's also happening on some London tiles immediately to the west of the
Greenwich Meridian.

I've had one further idea on this. Given the small values for longitude
at this point I wonder whether they're getting represented using
scientific notation at some point in the process, which may be confusing
some point further down the line that's trying to parse those strings
back as numbers?

Unfortunately I'm not really familiar enough with the t at h code to find
the error itself at the moment.

Gregory