[GRASS-dev] Re: New qgis-grass win32 build

[Sören Gebbert]

On Friday 05 May 2006 12:33, Radim Blazek wrote:
> On 5/5/06, "Sören Gebbert" <soerengebbert at gmx.de> wrote:
> > Dear Radim,
> > good news! :)
> >
> > The lates version of the grass test suite (0.2.0.2) is working
> > (mostly complete) within the
> > MinGW environment of your latest Qgis-GRASS windows build.
> > Only my output generation script
> > has some problems, but this is related to me.
> >
> > I made some screenshots of the xterm output of the test suite
> > for database, general, raster, raster3d and vector tests.
> 
> Nice and the result is not so bad. If you get the output working an HTML
> report will be available?

Of course .. but im still searching for the problem.

> 
> Did you check the buffer output with validation-failure? Does it mean
> that it works on Linux and fails on Windows? Is it the output realy wrong
> or maybe just a difference in non significant decimal places?

No, i only made the test runs. I did not verify the validation erros.
But the module v.buffer is kind of strange while generating output with many lines and 
buffers greater then ~50m on different machines (linux). 
I checkded this some time ago and figured out, that the output is equal, but
the order of the produced lines is different between the machines! I have no clue why.... .
In this case the md5 checksum test will fail. 
Even if i use valgrind for memchecks on the same machine, the v.buffer module shows this behavior:

You may want to  look at the v.buffer test part of:
http://www-pool.math.tu-berlin.de/~soeren/grass/GRASS_TestSuite/html/

The Normal precision difference between machines is catched, because the float 
output is truncatd of 3 decimal places via v.out.ascii.

> What does it mean dependencies not fulfilled?

This is a new feature. The test suite is trying to figure out what dependences the grass installation has.
e.g.: if gdal, ogr, postgresql, mysql and stuff is linked against grass

Within the test scripts you can define this dependences.
The r.in.gdal and r.out.gdal modules depend on gdal,
so i made the entry NEED_GDAL="yes" in the test scripts.

Example script: 
http://www-pool.math.tu-berlin.de/~soeren/grass/GRASS_TestSuite/html//r.out.gdal-test.sh.html

Bevor the test is executed, the test suite trys to figure out if gdal is installed by testing if r.in.gdal was build ..... very simple,
but i have no other way to figure this out (If the build information is written to a txt file which will be available like the BULD file in $GISBASE/etc,
things would be much easier for me. )
If r.in.gdal is not found, the test suite guess that gdal is not installed and linked, so every test depending on gdal is not executed
and a warning message is printed to stdout.

These are the dependences the test suite figured out within windows:

BLAS       no
CXX        no
DWG        no
FFMPEG     no
FFTW       no
FREETYPE   no
GDAL       no
GLW        no
JPEG       no
LAPACK     no
LFS        no
MOTIF      no
MYSQL      no
NLS        no
ODBC       no
OGR        no
OPENGL     no
PNG        no
POSTGRESQL yes
PROJ.4     no
READLINE   no
SQLITE     yes
TCLTK      no
TIFF       no
X11        no

Some of them like BLAS, LAPACK, READLINE, LFS and FREETYPE are not checked, because id dont know how!
But the values can be changes in the GlobalSettings script:
http://www-pool.math.tu-berlin.de/~soeren/grass/GRASS_TestSuite/html//GlobalSettings.html

Within the test scipts several dependences can be defined, you may want to look at the Template:
http://www-pool.math.tu-berlin.de/~soeren/grass/GRASS_TestSuite/html//TestScriptTemplate.html

Best regards
Soeren

btw.:
Can i run those tests within Windows 98? So i can use vmware
and dont have to boot Windows XP. I have only one XP license.

> 
> Radim
> 
> > Great work Radim!!!
> >
> > Best regards
> > Soeren
> >
> > btw.:
> > Sorry for those ugly pictures,
> > but my windows-gimp is not in a recent version.
> >
> > --
> > Echte DSL-Flatrate dauerhaft für 0,- Euro*!
> > "Feel free" mit GMX DSL! http://www.gmx.net/de/go/dsl
> >
> >
> 
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/grass-dev/attachments/20060505/20257679/attachment.html