On Wed, Dec 15, 2010 at 5:24 AM, Chris66 <[email protected]> wrote:
> Am 15.12.2010 11:48, schrieb Chris66:
>
>> Some more tests:
>>
>> NO effect (same sized output file) have:
>>
>> java u23 (32 Bit)
>> -Xmx1024 setting
>> -XX:-ReduceInitialCardMarks (gives 'unknown VM-option' error)
>> osmosis V0.37
>
> Interesting result when cutting the input file to below the
> 4 Gig limit:
>
> \cygwin\bin\head --bytes=3000000K europe.osm.pbf > test.osm.pbf
>
> Now osmosis it not stopping at 3.048.916.433.
>
> Output file has 8 Gig (still running).

Excellent experiment! I think you've found the cause.

The input file is just over 4gb. There's a 32-bit problem with the
end-of-file is, and thinks the file ends at around 4,457,858,847-2**32
=  162,891,551 bytes.

If I truncate off the first 162,891,551 bytes off of a geofabrik
europe extract and try to decode, I get 17,215,971 nodes, the exact
same as you reported with the broken file.

# head -c 162891551 /mnt/map/europe-created-on-linux.osm.pbf >
/tmp/foo.pbf ; ~/source/Map2/osmosis/package/bin/osmosis --read-pbf
/tmp/foo.pbf --write-xml file=- | wc
17215971 156188537 2524181063

Next question, why didn't this problem crop up earlier? Did a Java
update go out recently? Anyone know Java well enough to diagnose it
further, like why this problem is windows-only?

Scott

_______________________________________________
osmosis-dev mailing list
[email protected]
http://lists.openstreetmap.org/listinfo/osmosis-dev

Reply via email to