[postgis-devel] [PostGIS] #703: measures test failure
PostGIS
trac at osgeo.org
Mon Dec 13 05:26:38 PST 2010
#703: measures test failure
---------------------+------------------------------------------------------
Reporter: strk | Owner: pramsey
Type: defect | Status: new
Priority: medium | Milestone: PostGIS 2.0.0
Component: postgis | Version: trunk
Keywords: |
---------------------+------------------------------------------------------
Comment(by nicklas):
Ok, here is what I have found so far
I have compared a 64 bit ubuntu 10.4 installation with a 32 bit 10.10
installation on different hardware.
Both is PostgreSQL 9.01
The line numbers refere to revision 6238
the query I have used to compare is.
{{{
select st_astext(st_shortestline(a,b)) from (
select geomfromtext('LINESTRING(-40 -20 ,10 5)') as a,
geomfromtext('LINESTRING(-10 20, 1 -2)') as b
) as foo;
}}}
Both systems seems to store a slightly wrong value in variable r on line
882 in measures.c
r=8.0000000000000004e-1 instead of 0.8
Then, the difference between the systems occur on line 914 and 915
On the 32 bit system
A->x+r*(B->x-A->x) gives 2.22044607925031308e-15
and the 64 bit system gives
0.00000000000000000e+00[[BR]]
A->x and B->x seems to be identical.
So both systems get a floating point error when saving variable r but on
64 bit system that error doesn't seem to be reflected in the result. It
disappears in the next calculation.
To me it seems like 2.22044607925031308e-15 is to small to handle in one
system but possible to handle in the other. Or is it just rounded by some
other reason.
I will try to do the same outside postgis/postgresql in a standalone
program to see if the differnce seems to be in the system itself. But I
have no time right now.
/Nicklas
--
Ticket URL: <http://trac.osgeo.org/postgis/ticket/703#comment:6>
PostGIS <http://trac.osgeo.org/postgis/>
The PostGIS Trac is used for bug, enhancement & task tracking, a user and developer wiki, and a view into the subversion code repository of PostGIS project.
More information about the postgis-devel
mailing list