[postgis-users] PostGIS test failures

[Nathan Widmyer]

The tests I wrote seem to work correctly, so maybe the problems are confined
to just the two capabilities (compound curve and curve poly).

There are a few little things, such as shell scripts in the regress dir that
identify the /bin/sh (bourne shell) as the interpreter which is normally a
symlink to /bin/bash (bourne again shell), but of course in Solaris the
bourne shell still exists and doesn't have pushd.


I think the tests have failed because of a difference of endianness when
comparing the WKB definitions.

One example:
PostGIS tests expected WKB:
0109000000020000000102000000030000009FE5797057376340E09398B1B2373BC05AAE0A165F0963409F6760A2493D3DC0DB6286DFB057634082D8A1B32F843EC0010200000004000000DB6286DFB057634082D8A1B32F843EC075B4E4D0C60C634031FA5D1A371540C0D7197CED9B636340A3CB59A7630A41C050F4A72AC0FB6240974769FCE3CF41C0
astext: COMPOUNDCURVE((153.72942375 -27.2175704,152.29285719
-29.23940482,154.74034096 -30.51635287),(154.74034096
-30.51635287,152.39926953 -32.16574411,155.11278414
-34.08116619,151.86720784 -35.62414508))

received this actual WKB:
000000000900000002000000000200000003406337577079E59FC03B37B2B19893E04063095F160AAE5AC03D3D49A260679F406357B0DF8662DBC03E842FB3A1D882000000000200000004406357B0DF8662DBC03E842FB3A1D88240630CC6D0E4B475C04015371A5DFA314063639BED7C19D7C0410A63A759CBA34062FBC02AA7F450C041CFE3FC694797
astext: COMPOUNDCURVE((153.72942375 -27.2175704,152.29285719
-29.23940482,154.74034096 -30.51635287),(154.74034096
-30.51635287,152.39926953 -32.16574411,155.11278414
-34.08116619,151.86720784 -35.62414508))

Note the first two bytes, 01 is little endian and 00 is big endian, which
represent byte order, and they are different.
However, the WKT definitions themselves are equivalent.

Perhaps the tests need to be written to force a particular endianness, or
maybe compare the WKT representations instead?


I've attached a tarball of the test logfiles that get generated, which
include the test diffs, for whoever to examine.


Nate

On Thu, Aug 27, 2009 at 3:46 PM, <lighthousej at gmail.com> wrote:

> I installed PostgreSQL 8.4.0 and compiled in PostGIS 1.4.0 (GEOS 3.1.1 and
> PROJ.4.6.1) for Solaris sparc machine, 32-bit binaries.
>
> PostGIS didn't like being compiled partly by Sun's C compiler, cc. This
> setting is stored within the PostgreSQL global makefile configuration.
> I kept getting errors about unknown switches, etc... I edited the global
> settings makefile with the GNU equivalent and everything compiled now.
>
> I was just able to get the tests to run using the "make check" command and
> only two tests failed, sql-mm-compoundcurve and sql-mm-curvepoly.
>
> Rather than dumping the results immediately, here's the rundown:
> I differ on lines 93-97 of sql-mm-compoundcurve where it's comparing five
> different polygons.
> On sql-mm-curvepoly, on lines 134-138, and lines 164 and 167 where more
> polygons are being compared.
>
> I tried googling for phrases here and there but I keep getting the commit
> messages,commit diffs,etc...
>
> Is there something I did incorrectly when I built PostGIS?
>
> Thanks,
> Nate
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20090828/d21a9a3d/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pgis_reg_test_solaris.tar.gz
Type: application/x-gzip
Size: 3632 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20090828/d21a9a3d/attachment.bin>