[Tile-serving] [osm2pgsql] Don't let invalid geometries into the database (#428)

[Paul Norman]

> My main point was that this is not a bug fix

I've been viewing invalid geoms as a bug, and this as a bug fix.

> Third point: st_isvalid is known to be slow, I don't know about geos' is_valid() function. Did you see any performance impact because of the additional valid checks?

It's only called a second time on a geom that is invalid before buffering. Unfortunately, these do tend to be the more complex geoms

master vs master + this PR merged does show a performance difference
```
            master  PR
proc rels   11343   12176
pend ways    6752    6811
other       14823   14820
total       32918   33807
```

That's an increase of 889s for a planet import, or 2.7%.

---
Reply to this email directly or view it on GitHub:
https://github.com/openstreetmap/osm2pgsql/pull/428#issuecomment-130177938
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/tile-serving/attachments/20150811/16ffab26/attachment.html>