[mapserver-users] Corrupt image output (solved?)

[Mike Leahy]

Hello again,

I think I found the cause of my problem - on my server (where I had 
preprocessed all of my imagery), I have libgeotiff-1.2.4 installed from 
the F8 repositories.  On my laptop I had libgeotiff-1.2.3 installed from 
source.  I upgraded my laptop to libgeotiff-1.2.4 (I tried 1.2.5, but it 
didn't compile for me), and recompiled gdal and mapserver.  Now it seems 
that imagery renders okay (after some quick tests).  I didn't realize I 
had different libgeotiff library versions...but it also hadn't occurred 
to me that a GeoTiff dataset might be incompatible with different 
versions of libgeotiff.

So should I maybe close the ticket that I opened earlier today?

Mike

Mike Leahy wrote:
> Ok, I'll file a ticket shortly...but I have a few more things I've tried 
> that might add insight to this.  It seems if I load the map via cgi 
> mapserver, I get the same problems - so this isn't limited to mapscript. 
>  In contrast, I don't seem to have any trouble with the output values if 
> I extract a single pixel from an image using gdal_translate to a *.grd 
> file (at the same coordinates where I get corrupt results from 
> mapscript/mapserver) - so the problem doesn't seem to be with gdal by 
> itself.
> 
> Also, I can point out that the corrupted areas in the output have no 
> coincidence with tile-index boundaries.  I say this out because if I 
> render a layer from a non-tiled image dataset, the corrupted areas 
> appear as horizontal bands across the entire width of the image output, 
> rather than rectangular areas (though my non-tiled dataset is a 
> single-band floating-point DEM, rather than an rgb image).  However, as 
> before, the height and number of lines of each band is variable.
> 
> Maybe it's just time to upgrade my OS (F7 is EOL after all).
> 
> Regards,
> Mike
> 
> Tamas Szekeres wrote:
>> Mike,
>>
>> It's quite difficult to find out what's going on. You should file a
>> ticket including the example and sample data. Probably the problem may
>> be reproduced in other machines as well.
>>
>> Best regards,
>>
>> Tamas
>>
>>
>>
>> 2008/8/19 Mike Leahy <mgleahy at alumni.uwaterloo.ca>:
>>> Tamas,
>>>
>>> I just gave this a try a few times - deleted the cache, and loaded 
>>> one tile.
>>>  It seems to happen less frequently, but I still have about a 1 in 3 
>>> chance
>>> of getting various rectangle artifacts on the image.  It basically looks
>>> like something forgets to draw one or more of the rgb colours.
>>>
>>> Mike
>>>
>>> Tamas Szekeres wrote:
>>>> Mike,
>>>>
>>>> Did you experience the same problem when calling tile.php directly
>>>> when rendering an individual tile?
>>>>
>>>> Best regards,
>>>>
>>>> Tamas
>>>>
>>>>
>>>> 2008/8/19 Mike Leahy <mgleahy at alumni.uwaterloo.ca>:
>>>>> Hello list,
>>>>>
>>>>> I'm wondering if anyone can help me figure out a problem I'm having 
>>>>> with
>>>>> corrupted output from from a tiled geotiff dataset I'm using in a 
>>>>> ka-Map
>>>>> environment.  You can see what's happening to the imagery here:
>>>>> http://img527.imageshack.us/my.php?image=corrupthd1.png
>>>>>
>>>>> This generally happens when I'm zoomed in fairly close, and usually 
>>>>> with
>>>>> this high-resolution image dataset.  However, I can get corrupted 
>>>>> output
>>>>> for
>>>>> other raster datasets (all geotiff, but not always tiled).  It 
>>>>> seems to
>>>>> happen when the machine is processing most intensively (i.e., usually
>>>>> when
>>>>> the imagery is being rendered by ka-Map).  It also seems to be 
>>>>> limited to
>>>>> the laptop I am using, which is running F7-x86_64.  In contrast, I 
>>>>> have
>>>>> an
>>>>> actual server (F8-x86_64), which has a virtually identical setup (with
>>>>> respect to MapServer and its various dependencies), but with lots more
>>>>> cpu
>>>>> cores and ram, and it's having no trouble at all (as far as I have
>>>>> noticed).
>>>>>
>>>>> The corruption in the output from the same application my laptop is 
>>>>> not
>>>>> always the same - if I clear the cache for ka-Map and re-render the
>>>>> images,
>>>>> I'll get similar artifacts appearing in the output (red, green, blue
>>>>> and/or
>>>>> black rectangles), but they will vary in size and location.  It 
>>>>> happens
>>>>> for
>>>>> both GD/AGG drivers, and JPEG/PNG output formats.
>>>>>
>>>>> Until now, I figured it was just a problem with limited memory on my
>>>>> laptop
>>>>> (it was always around 100% when rendering this imagery on the web 
>>>>> service
>>>>> +
>>>>> viewing the application in a browser), but now I've upgraded the ram -
>>>>> there's plenty of spare room and I'm still getting the same 
>>>>> problem.  I
>>>>> just
>>>>> finished upgrading Gdal from 1.5.0 to 1.5.2, and that didn't help any.
>>>>>
>>>>> Basically, the only real differences I can think of between my 
>>>>> laptop and
>>>>> my
>>>>> server are the number of cpu cores (1 on my laptop vs. 8 on the 
>>>>> server),
>>>>> and
>>>>> possibly the versions of some of the underlying libraries in the 
>>>>> system.
>>>>>  Does anyone know where I should start looking to find the cause of 
>>>>> this?
>>>>>  Or perhaps somebody recognizes this problem, and knows what I 
>>>>> might be
>>>>> doing wrong...
>>>>>
>>>>> Thanks for any suggestions,
>>>>> Mike
>>>>>
>>>>> _______________________________________________
>>>>> mapserver-users mailing list
>>>>> mapserver-users at lists.osgeo.org
>>>>> http://lists.osgeo.org/mailman/listinfo/mapserver-users
>>>>>
>>
>