Recent CVS segfaults, msautotest and python unit tests

[Frank Warmerdam]

Umberto Nicoletti wrote:
> Now I am facing another problem: the msautotest suite reports a large
> number of failures (results that don't match). Comparing each couple
> of images (result and expected) doesn't help because the slight, but
> well visible differences don't seem due to my modifications anyway.
> Can someone with more knowledge of msautotest than I have shed some
> light and confirm whether I am on the good track or not?

Umberto,

Unfortunately the exact images produced by MapServer are very
subject to details of floating point round off, exact version of
GD, and other variables.  So the msautotest suite almost always
produces many differences, and it is necessary to do visual inspections
of them to verify they are at least reasonable.

I'd love to find some approach that produces exact image results on
different platforms!

In the meantime, I'll do a review current msautotest results with your
changes and verify they seem ok to me.

> Last, but not least, running msautotest and the python unit tests is
> not possible under python 2.5. To fix the former simply apply this
> patch (4th argument to cmp is deprecated since python 2.3):
> 
> Index: mstestlib.py
> ===================================================================
> RCS file: /data2/cvsroot/msautotest/pymod/mstestlib.py,v
> retrieving revision 1.11
> diff -r1.11 mstestlib.py
> 113c113
> <     if filecmp.cmp(expected_file,result_file,0,1):
> ---
>>     if filecmp.cmp(expected_file,result_file,0):
> 319a320,321
>>           #print command
>>

I have reviewed the old docs, and it doesn't seem this use_statcache
option was ever really required.  I'll remove it in trunk and 4.10
branch.

Best regards,
-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | President OSGeo, http://osgeo.org