On Oct 12, 2011, at 9:51 AM, Iván Sánchez Ortega wrote:

> Hi all,
> 
> After seeing it in action, I'm jumping to try out tilemill and carto first-
> hand. But I'm running into several problems...
> 
> Environment: I run Debian testing/unstable, amd64 architecture. Boost 1.46, 
> Python 2.7, Mapnik 2.0.0, Node.js 0.4.12, the works.
> 

That stack should work absolutely fine. I'm assuming that your Mapnik 2.0.0 
install is from source? (or at least you installed it yourself?) It should be.

> On the tilemill front: I'm able to install it without too much fuzz (dpkg --
> force-all does wonders),

Do not use the TileMill debian package if you have installed Mapnik yourself 
from source.

The reason is that the TileMill .deb pulls in its own .deb of mapnik which will 
conflict with any source installed version.

The expected outcome of conflicting mapnik versions is that tilemill's server 
will segfault (really node-mapnik, the js bindings to mapnik will segfault).

> I'm able to start it up, and I'm able to point a web 
> browser to localhost:8889. But as soon as the browser requests a tile, the 
> tilemill server process segfaults without telling where or why.
> 
> Question: How can I increase the verbosity of tilemill so I can trace the 
> problem?

You would need to run tilemill in gbd and then you would see a crash somewhere 
in node-mapnik.

So, the solution to this will not be more verbose output but to uninstall the 
TileMill .deb and re-install TileMill from *source* and in doing so it will 
leverage your existing installation of Mapnik 2.x

The instructions for Ubuntu should work fine for debian and are at: 
http://mapbox.com/tilemill/docs/source/

> On the carto front, javascript version. I can do "aptitude install npm" and 
> then "npm install carto" without any problems. But when I try run it, I only 
> got this:
> 
> $ carto /usr/share/tilemill/examples/control-room/control-room.mml 
> carto: Millstone not found. Externals will not be resolved.

Do also:

npm install millstone

> 
> On the carto-generator front, C version. When trying to compile it, it 
> ultimately fails with:
> 
> /usr/include/boost/spirit/home/karma/numeric/detail/real_utils.hpp:98:64: 
> error: invalid cast from type 
> ‘boost::fusion::extension::adt_attribute_proxy<mapnik::building_symbolizer, 
> 2, 
> true>’ to type ‘double’

The C++ version requires at least boost 1.47. We plan to move this code into 
Mapnik core soon, and once we do this a configure time check will be able to 
warn you about this restriction.

_______________________________________________
Mapnik-users mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/mapnik-users

Reply via email to