[postgis-devel] Fwd: [postgis-users] 1.3.4rc2 Released

[Paul Ramsey]

Curve fixes could be a while....


---------- Forwarded message ----------
From: Mark Leslie <mark.leslie at lisasoft.com>
Date: Thu, Nov 6, 2008 at 5:16 PM
Subject: Re: [postgis-users] 1.3.4rc2 Released
To: Havard Tveite <havard.tveite at umb.no>
Cc: Paul Ramsey <pramsey at cleverelephant.ca>


I'm sad to report that I really haven't made any progress on this.
The sun box doesn't seem to be reporting anything from the box2d/3d
calls, but is passing the simpler ones.  I'm still trying to get
access to our only sparc machine in the office, but they need to clean
it up some before I'm allowed to break it again.

The only difference I'm aware of between the architectures is the
floating point error, generally below the 1e-8 to 1e-12 level.  This
is not the case in the results you're seeing, but it is conceivable
that an internal comparison is getting botched due to this.  Looking
at the intermediate values, I think this is very unlikely, but I have
gone through the algorithms and replaced all the floating point
comparisons with the clever macros setup for that purpose (branch
1.3).  Why I didn't use those in the first place is beyond me.

I'm on leave now until December, and won't have net access until I
return.  Hopefully elephants and rain will inspire me.

Mark Leslie
Geospatial Software Architect
LISAsoft

-------------------------------------------------------------
Ph: +61 2 8570 5000 Fax: +61 2 8570 5099 Mob: +61
Suite 112, Jones Bay Wharf 19-21 Pirrama Rd Pyrmont NSW 2009
-------------------------------------------------------------

LISAsoft is part of the A2end Group of Companies
http://www.ardec.com.au
http://www.lisasoft.com
http://www.terrapages.com


Havard Tveite wrote:
>
> Dear Mark,
>
> I assume you wanted me to run the test with current SVN.
> I attach the new diff file.
>
> What I have done:
> 1) Modified lwgeom/liblwgeom.h to set PGIS_DEBUG = 1 and
>    PGIS_DEBUG_CALLS = 1
> 2) ./configure
> 3) make
> 4) edited regress/lwpostgis.sql to reference my build
>    directory instead of $libdir
> 5) edited regress/Makefile to run only sql-mm-circularstring
> 6) make check
>
> Output:
> -------------------------------------------------------
> $ make check
> make -C regress test
> make[1]: Entering directory `/data3/usr/local/src/postgresql/postgresql-8.1.5/contrib/postgis-1.3.4svn/regress'
> Creating spatial db postgis_reg
> TMPDIR is /tmp/pgis_reg_1636
>
>  PostgreSQL 8.1.5 on sparc-sun-solaris2.7, compiled by GCC gcc (GCC) 3.3.2
>  Postgis 1.3.4SVN - 2008-11-04 08:29:31
>    GEOS: 3.0.0-CAPI-1.4.1
>    PROJ: Rel. 4.5.0, 22 Oct 2006
>
> Running tests
>
>  sql-mm-circularstring. failed (diff expected obtained: /tmp/pgis_reg_1636/test_1_diff)
>
> Run tests: 1
> Failed: 1
> -------------------------------------------------------
>
> Håvard
>
>
> Mark Leslie wrote:
>>
>> I've just committed changes to the test that will run the box2d/3d and
>> check the results rounded to 1e-8.  This is our default floating point
>> tolerance, at least in the curved arena, but it may be too small for the
>> box2d (4byte) values.  At any rate, can you run just the
>> sql-mm-circularstring test again, with a clean build and send me the
>> test diff?
>>
>> I am fleeing the country for the rest of the month, starting at the end
>> of the week, so I'll try to get this sorted by then.  Otherwise I won't
>> have computer access until December.  Is this the last issue we're
>> waiting on for the 1.3.4 release?
>>
>> --
>> Mark
>>
>> Mark Leslie wrote:
>>>
>>> No excuses necessary, I was very unclear.
>>> The file you sent is the correct one, but it doesn't contain nearly as
>>> much output as I would expect.  Nothing on the lwcurve.c functions at
>>> all from what I've seen.  Have you run a 'make clean' after changing the
>>> macros?
>>> You can run this more easily with : psql -c "SELECT
>>> ST_box2d('CIRCULARSTRING(220268.439465645
>>> 150415.359530563,220227.333322076 150505.561285879,220227.353105332
>>> 150406.434743975)'::geometry)" template-postgis &> output
>>>
>>> Looking through the regression test, I see that I still have the
>>> previous ST_box2d/3d tests commented out, so it looks like I've never
>>> gotten this sorted out properly.  I'm looking at how to get all the
>>> tests back in, so hopefully we'll have nine failing calls instead of
>>> one.  But that will give me a bit more to grab onto.
>>>
>>> Havard Tveite wrote:
>>>
>>>> Please excuse my ignorance.  I am not at all familiar with
>>>> the regression tests.
>>>>
>>>> I have now tried to set PGIS_DEBUG = 1 and PGIS_DEBUG_CALLS = 1.
>>>> Running make check became a quite tough task for my old
>>>> machine - I stopped it when it had run for some hours and had
>>>> generated over 400 MB of output for test_14_tmp.
>>>>
>>>> I tried again, removing all tests except sql-mm-circularstring
>>>>
>>>> Here is the output:
>>>> *************************************
>>>> $ make check
>>>> make -C regress test
>>>> make[1]: Entering directory `/data3/usr/local/src/postgresql/postgresql-8.1.5/contrib/postgis-1.3.4rc2/regress'
>>>> Creating spatial db postgis_reg
>>>> TMPDIR is /tmp/pgis_reg_15236
>>>>
>>>>  PostgreSQL 8.1.5 on sparc-sun-solaris2.7, compiled by GCC gcc (GCC) 3.3.2
>>>>  Postgis 1.3.4rc2 - 2008-10-31 08:57:52
>>>>    GEOS: 3.0.0-CAPI-1.4.1
>>>>    PROJ: Rel. 4.5.0, 22 Oct 2006
>>>>
>>>> Running tests
>>>>
>>>>  sql-mm-circularstring. failed (diff expected obtained: /tmp/pgis_reg_15236/test_1_diff)
>>>>
>>>> Run tests: 1
>>>> Failed: 1
>>>> *************************************
>>>>
>>>> I attach /tmp/pgis_reg_15236/test_1_diff
>>>>
>>>> I hope that all you need is in there - or do you need something
>>>> from the PostgreSQL log file also?
>>>>
>>>> Håvard
>>>>
>>>>
>>>>
>>>> Paul Ramsey wrote:
>>>>
>>>>
>>
>