Opened 16 years ago
Closed 9 years ago
#2260 closed defect (fixed)
stripes in ArcInfo binary grid
Reported by: | bartvde | Owned by: | warmerdam |
---|---|---|---|
Priority: | normal | Milestone: | 1.8.1 |
Component: | GDAL_Raster | Version: | 1.5.0 |
Severity: | normal | Keywords: | HFA, TIF |
Cc: | Sam Gillingham |
Description
art van den Eijnden (OSGIS) wrote:
Hi list,
after upgrading to GDAL 1.5 (coming from 1.4) we have some stripes in an ArcInfo binary grid (viewed with Mapserver).
Any ideas what could be causing this?
The problem can be seen in this screendump:
Bart,
If you can file a ticket on this, and provide a pointer to the problem dataset I'll treat this as priority to fix before the soon to be released GDAL 1.5.1.
There have been some improvements to the tile decoder for arc/info grids for 1.5.0 but with no specifications and only a haphazard assortment of test data it is hard to be sure one improvement doesn't break some other case.
Best regards,
Change History (9)
comment:1 by , 16 years ago
Component: | default → GDAL_Raster |
---|
comment:2 by , 16 years ago
comment:3 by , 16 years ago
Keywords: | HFA added |
---|---|
Milestone: | → 1.5.1 |
Priority: | normal → highest |
Status: | new → assigned |
It turns out the problem is actually the HFA driver reading the .rrd overviews, not the AIGRID driver. A temporary work around would be to delete the .aux and .rrd file and regenerate overviews with gdaladdo.
Digging deeper into the HFA problem.
comment:4 by , 16 years ago
Cc: | added |
---|---|
Milestone: | 1.5.1 → 1.5.2 |
Priority: | highest → normal |
I have reviewed a problem tile from the .rrd file. The tile is 256x4 and starts at offset 0x347B and is 145 bytes long (to 0x350C).
00003470 01 00 00 00 0f 00 00 00 00 44 00 00 00 00 00 80 |.........D......| 00003480 00 00 00 91 00 00 00 00 40 e1 01 01 01 01 01 01 |........@.......| 00003490 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 |................| 000034a0 01 01 01 01 01 01 01 01 01 40 e1 01 01 01 01 01 |.........@......| 000034b0 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 |................| 000034c0 01 01 01 01 01 01 01 01 01 01 40 e1 01 01 01 01 |..........@.....| 000034d0 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 |................| 000034e0 01 01 01 01 01 01 01 01 01 01 01 40 e1 01 01 01 |...........@....| 000034f0 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 |................| 00003500 01 01 01 01 01 01 01 01 01 01 01 01 05 00 00 00 |................|
The UncompressBlock() decodes this as nMinVal=0, nNumRuns=128, nDataOffset=145 and nNumBits=0. This decodes as 1024 pixels with value zero *but* clearly some of those pixels are intended to be the proper min-val for the dataset which is -2147483647 (-MIN_INT). I have been unable to deduce how this tiled can actually be decoded. The current code behavior seems correct for the data available even though it clearly doesn't produce the expected result.
I looked back in the code and I can't see any obvious way this could ever have worked. However, r11142 for #1000 likely changed the behavior from returning the nodata value for all pixels to returning zeros for all pixels. Perhaps it wasn't really working before, but it was less obvious?
Adding Sam to the ticket since he was involved in the earlier changes and might have some wisdom to offer.
Also pushing off to 1.5.2 and reducing priority since I'm not sure it is solvable, and it isn't clear this really represents a regression compared to 1.4.x.
comment:6 by , 16 years ago
The more I look into these problems the less I think we understand the format completely. Unfortunately I don't seem to be getting anywhere with Leica in regards to the format specification (see #1000). I'll see if I can work out another angle on them.
comment:7 by , 16 years ago
Keywords: | TIF added |
---|
Hi All, we are having typically the same trouble here. But we don't use Arcinfo type.
There's color stripes on rendered images. You will find the image issue from mapserver.
This appear on our server from the update of gdal from 1.4x to 1.5x (actually 1.5.1 rpm made by the openbuilder opensuse website). we are on 64bits servers.
http://linux.ioda.net/images/n16_defect.jpg
This tend to prove that it's not probably a trouble attached to arcinfo.
comment:8 by , 9 years ago
bartvde, do you believe that this issue is still possible to reproduce with reasonable pain? Or would it be better to close the ticket?
comment:9 by , 9 years ago
Resolution: | → fixed |
---|---|
Status: | assigned → closed |
this is so old, let's close it ;-)
Mapserver LAYER definition: