[osmosis-dev] pbf problems under windows

[Frederik Ramm]

Hi,

Scott Crosby wrote:
> If they do that conversion on windows, does that generate a pbf that
> has the same md5sum as the geofabrik generated file? (And if not, then
> I'd like both of those files!)

I'll try to assemble a proper test case with the help of one of the 
windows users.

>> * converting the Geofabrik-generated .osm.pbf to XML will copy exactly
>> 10,600,000 nodes to the XML file and then stop, without error, adding a
>> proper "</osm>" but no ways or relations.
> 
> Is this with pbf2osm or osmosis or both?

This is when trying an Osmosis-on-Linux-generated .osm.pbf to .osm with 
Osmosis-on-Windows. pbf2osm is not part of this particular problem.

> Is the 10,600,000 nodes from different files or the same file?

The resulting XML file has 10,600,000 nodes.

>> Does anybody have an idea why this might be? The europe.osm.pbf can be
>> processed without problems on Linux, with Osmsis as well as with the mkgmap
>> splitter.
> 
> Did the users having this problem successfully work with geofabrik
> extracts from last week or last month on their windows machines?

Yes, they claim they did. The only change that I made is upgrading from 
0.37 to latest SVN, and adding the cascadingRelations=true option which 
should really only influence relations.

> Other than that, bizarre. I had some random crashes with the mkgmap
> splitter, until I found this bug:
> http://bugs.sun.com/view_bug.do?bug_id=6888898  and started running
> with -XX:-ReduceInitialCardMarks
> 
> I'm wondering if this is some issue with corruption with the JDK
> that's rare and causes spurrious correlations. That would explain why
> smaller files are less likely to hit the issue, whatever it is.

I think one of the error reporters indeed seems to have some sort of 
hardware (RAM) fault but it is unlikely that all of them do.

> I don't have development environment on windows to do any experiments of my own.

Same here although I might be able to conjure something up on an old laptop.

Bye
Frederik

-- 
Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00'09" E008°23'33"