|
Hi,
I have tried to use GDAL 2.2 in OTB. I have some issues with
output images in the ENVI format. Some of the images I use have :
- A geotransform with an origin, for instance : geotransform =
{10, 1, 0, 10, 0, 1}
- No map projection (the image comes from a sensor image, or may
come from a plain jpg file).
As you can see, the 2x2 linear part of the geotransform is an
identity matrix. When I write such an image using the ENVI format, I
get a 90° rotation :
map info = {Arbitrary, 1, 1, 10, 10, 1, 1, 0, North,
rotation=90}
Does ENVI support negative values for "Pixel Y size", in order to
handle this Y-flipping ?
Regards,
Guillaume Pasero
On 03/01/2017 04:03 PM, Kurt Schwehr
wrote:
Even,
I already do direct tests with jp2kakdataset_test.cc with
bazel and it works great. I can directly probe the methods
without worrying about any other mechanism getting between the
test and the class. e.g. I can directly mess with Identify,
Open and any helpers. I can also do things like disable
identify so that fuzzing can more quickly get into the guts
and find trouble:
GDALDataset *JP2KAKDataset::Open( GDALOpenInfo *
poOpenInfo ) {
#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
// During fuzzing, do not use Identify to reject
crazy content.
if( !Identify(poOpenInfo) )
return NULL;
#endif
We are putting resources behind this... we are running
fuzzing on jp2kak and related. e.g. with >600K core hours
of fuzzing this month (> 14e12 iterations).
If I don't break things apart, fuzzing a component goes
much slower and is far less likely to identify issues in the
lower level components and libs.
|
_______________________________________________
gdal-dev mailing list
gdal-dev@lists.osgeo.org
https://lists.osgeo.org/mailman/listinfo/gdal-dev
